Gemini CLI

https://blog.google/technology/developers/introducing-gemini-cli-open-source-ai-agent/

by sync • 1 year ago

1428 788 1 year ago

blog.google

GitHub: https://github.com/google-gemini/gemini-cli

thor-rodrigues• 1 year ago

Link of announcement blog post: https://blog.google/technology/developers/introducing-gemini...

sync2• 1 year ago

These always contain easter eggs. I got some swag from Claude Code, and as suspected, Gemini CLI includes `/corgi` to activate corgi mode.

GavCo• 1 year ago

They sent you swag in the mail? How did that work?

sync2• 1 year ago

Yeah, I'm not sure if it's still there (their source code is increasingly obfuscated) but if you check out the source for the first public version (0.2.9) you'll see the following:

    Sends the user swag stickers with love from Anthropic.",bq2=`This tool should be used whenever a user expresses interest in receiving Anthropic or Claude stickers, swag, or merchandise. When triggered, it will display a shipping form for the user to enter their mailing address and contact details. Once submitted, Anthropic will process the request and ship stickers to the provided address.
    
    Common trigger phrases to watch for:
    - "Can I get some Anthropic stickers please?"
    - "How do I get Anthropic swag?"
    - "I'd love some Claude stickers"
    - "Where can I get merchandise?"
    - Any mention of wanting stickers or swag
    
    The tool handles the entire request process by showing an interactive form to collect shipping information.

ed_mercer• 1 year ago

That’s why we’re introducing Gemini CLI

Definitely not because of Claude Code eating our lunch!

troupo• 1 year ago

And since they have essentially unlimited money they can offer a lot for free/cheaply, until all competitors die out, and then they can crank up the prices

pzo• 1 year ago

yeah we already seen this with gemini 2.5 flash. Gemini 2.0 is such a work horse for API model with great price. Gemini 2.5 flash lite same price but is not as good except math and coding (very niche use case for API key)

unshavedyak3• 1 year ago

Yea, i'm not even really interested in Gemini atm because last i tried 2.5 Pro it was really difficult to shape behavior. It would be too wordy, or offer too many comments, etc - i couldn't seem to change some base behaviors, get it to focus on just one thing.

Which is surprising because at first i was ready to re-up my Google life. I've been very anti-Google for ages, but at first 2.5 Pro looked so good that i felt it was a huge winner. It just wasn't enjoyable to use because i was often at war with it.

Sonnet/Opus via Claude Code are definitely less intelligent than my early tests of 2.5 Pro, but they're reasonable, listen, stay on task and etc.

I'm sure i'll retry eventually though. Though the subscription complexity with Gemini sounds annoying.

ur-whale4• 1 year ago

It would be too wordy, or offer too many comments

Wholeheartedly agree.

Both when chatting in text mode or when asking it to produce code.

The verbosity of the code is the worse. Comments often longer than the actual code, every nook and cranny of an algorithm unrolled over 100's of lines, most of which unnecessary.

Feels like typical code a mediocre Java developer would produce in the early 2000's

sirn• 1 year ago

I've found that Gemini 2.5 Pro is pretty good at analyzing existing code, but really bad at generating a new code. When I use Gemini with Aider, my session usually went like:

    Me: build a plan to build X
    Gemini: I'll do A, B, and C to achieve X
    Me: that sounds really good, please do
    Gemini: <do A, D, E>
    Me: no, please do B and C.
    Gemini: I apologize. <do A', C, F>
    Me: no! A was already correct, please revert. Also do B and C.
    Gemini: <revert the code to A, D, E>

Whereas Sonnet/Opus on average took me more tries to get it to the implementation plan that I'm satisfied with, but it's so much easier to steer to make it produce the code that I want.

jstummbillig2• 1 year ago

I find it hard to imagine that any of the major model vendors are suffering from demand shortages right now (if that's what you mean?)

If you mean: This is "inspired" by the success of Claude Code. Sure, I guess, but it's also not like Claude Code brought anything entirely new to the table. There is a lot of copying from each other and continually improving upon that, and it's great for the users and model providers alike.

coolKid721• 1 year ago

ai power users will drop shit immediately, yes they probably have long term contracts with companies but anyone seriously engaged has switched to claude code now (probably including many devs AT openai/google/etc.)

If you don't think claude code is just miles ahead of other things you haven't been using it (or well)

I am certain they keep metrics on those "power users" (especially since they probably work there) and when everyone drops what they were using and moves to a specific tool that is something they should be careful of.

wiseowise7• 1 year ago

I will never understand this kind of snark comment.

What are they supposed to do?

“Oh no, they’ve released CLI tool before us! It’s game over, we can’t do it too, we need to come up with something else now!”

htrp• 1 year ago

symptomatic of Google's lack of innovation and pm's rushing to copy competitor products

better question is why do you need a modle specific CLI when you should be able to plug in to individual models.

jvanderbot• 1 year ago

Aider is what you want for that.

wagslane• 1 year ago

check out opencode by sst

shmoogy• 1 year ago

If Claude code is any indication it's because they can tweak it and dogfood to extract maximum performance from it. I strongly prefer Claude code to aider - irrespective of the max plan.

Haven't used Jules or codex yet since I've been happy and am working on optimizing my current workflow

poszlem• 1 year ago

The killer feature of Claude Code is that you can just pay for Max and not worry about API billing. It lets me use it pretty much all the time without stressing over every penny or checking the billing page. Until they do that - I'm sticking with Claude.

mhb• 1 year ago

How does that compare to using aider with Claude models?

adamcharnock2• 1 year ago

I did a little digging into this just yesterday. The impression I got was that Claude Code was pretty great, but also used a _lot_ more tokens than similar work using aider. Conversations I saw stated 5-10x more.

So yes with Claude Code you can grab the Max plan and not worry too much about usage. With Aider you'll be paying per API call, but it will cost quite a bit less than the similar work if using Claude Code in API-mode.

I concluded that – for me – Claude Code _may_ give me better results, but Aider will likely be cheaper than Claude Code in either API-mode or subscription-mode. Also I like that I really can fill up the aider context window if I want to, and I'm in control of that.

therealmarv2• 1 year ago

Using Claude models in aider burns tokens you need to top up. With Claude Max subscription you can pay a 100 or 200 USD per month plan and use their internal tool claude code without the need to buy additional pay as you go tokens. You get a "flatrate", the higher plan gives you more usage with less rate limiting.

Karrot_Kream• 1 year ago

Aider and Claude Code/Gemini CLI agentic stuff operate differently.

You can think of Aider as being a semi-auto LLM process. First you ask it to do something. It goes through a generate -> reflect -> refine loop until it feels like it has achieved the goal you give it. Aider has a reflection limit so it'll only do this loop a limited number of times and then it will add/remove the code that it deems fit. Then it'll give you instructions to run. You can run those instructions (e.g. to actually run a script) and then append the results from the run into the context to get it to fix any issues, but this is optional. What you send in the context and what you ask the models to do are in your hands. This makes iteration slower and the LLM does less but it also can potentially keep costs lower depending on what you delegate to the LLM and how often you iterate with it.

Claude Code, Codex, and I suspect Gemini CLI on the other hand will autonomously run your code then use the output to continue refining its approach autonomously until the goal is reached. This can consume many more tokens, potentially, than hand guiding Aider, because its potential for iteration is so much longer. But Claude Code and the like also need a lot less direction to make progress. You can, for example, ask it to do a big refactor and then just leave to lunch and come back to see if the refactor is done. Aider will require babying the whole way.

therealmarv2• 1 year ago

That's a golden cage and you limit yourself to Anthropic only.

I'm happy I can switch models as I like with Aider. The top models from different companies see different things in my experiences and have their own strengths and weaknesses. I also do not see Anthropic's models on the top of my (subjective) list.

jedi3335• 1 year ago

No per-token billing here either: "...we offer the industry’s largest allowance: 60 model requests per minute and 1,000 requests per day at no charge."

https://blog.google/technology/developers/introducing-gemini...

thimabi15• 1 year ago

Don’t know about Claude, but usually Google’s free offers have no privacy protections whatsoever — all data is kept and used for training purposes, including manual human review.

unshavedyak3• 1 year ago

Same. Generally i really prefer Claude Code's UX (CLI based, permissions, etc) - it's all generally close to right for me, but not perfect.

However i didn't use Claude Code before the Max plan because i just fret about some untrusted AI going ham on some stupid logic and burning credits.

If it's dumb on Max i don't mind, just some time wasted. If it's dumb on credits, i just paid for throw away work. Mentally it's just too much overhead for me as i end up worrying about Claude's journey, not just the destination. And the journey is often really bad, even for Claude.

rusk• 1 year ago

This insistence by SAAS vendors upon not protecting you from financial ruin must surely be some sort of deadweight loss.

Sure you might make a few quick wins from careless users but overall it creates an environment of distrust where users are watching their pennies and lots are even just standing off.

I can accept that with all the different moving parts this may be a trickier problem than a pre paid pump, or even a Telco, and while to a product manager this might look like a lot of work/money for something that “prevents” users overspending.

But we all know that’s shortsighted and stupid and its the kind of thinking that broadly signals more competition is required.

fhinkel17• 1 year ago

If you use your personal gmail account without billing enabled, you get generous requests and never have to worry about a surprise bill.

indigodaddy5• 1 year ago

If I have a CC linked to my personal Google for my Google One storage and YouTube Premium, that doesn't make me "billing enabled" for Gemini CLI does it?

lherron2• 1 year ago

Hope this will pressure Anthropic into releasing Claude Code as open source.

zackify2• 1 year ago

What’s neat is we can proxy requests from Gemini or fork it with only replacing the api call layer so it can be used with local models!!!

willsmith723• 1 year ago

As a heavy Claude code user that's not really a selling point for me

Ultimately quality wins out with LLMs. Having switched a lot between openai, google and Claude, I feel there's essentially 0 switching cost and you very quickly get to feel which is the best. So until Claude has a solid competitor I'll use it, open source or not

lherron2• 1 year ago

Even if you don't care about open source, you should care about all the obfuscation happening in the prompts/models being used by Cursor/Claude Code/etc. With everything hidden, you could be paying 200/mo and get served Haiku instead of Sonnet/Opus. Or you could be getting 1k tokens of your code inserted as context instead of 100k to save on inference costs.

fhinkel17• 1 year ago

I love healthy competition that leads to better use experiences

asadm10• 1 year ago

I have been using this for about a month and it’s a beast, mostly thanks to 2.5pro being SOTA and also how it leverages that huge 1M context window. Other tools either preemptively compress context or try to read files partially.

I have thrown very large codebases at this and it has been able to navigate and learn them effortlessly.

zackify2• 1 year ago

When I was using it in cursor recently, I found it would break imports in large python files. Claude never did this. Do you have any weird issues using Gemini? I’m excited to try the cli today

asadm10• 1 year ago

not at all. these new models mostly write compiling code.

_zoltan_• 1 year ago

what's your workflow?

leoh7• 1 year ago

+1 I have not found Gemini 2.5 better than Claude's latest models -- different and better at some things, but not better in general; and in general I have found Gemini 2.5 Pro to be worse at dealing with large codebases despite its huge context window. So I too am quite curious about the workflow here.

asadm10• 1 year ago

i guess one core idea is to point it to an entrypoint or some part of code you are planning to touch. instead of asking it to load up every file etc. and then figure out the control flow.

iandanforth2• 1 year ago

I love how fragmented Google's Gemini offerings are. I'm a Pro subscriber, but I now learn I should be a "Gemini Code Assist Standard or Enterprise" user to get additional usage. I didn't even know that existed! As a run of the mill Google user I get a generous usage tier but paying them specifically for "Gemini" doesn't get me anything when it comes to "Gemini CLI". Delightful!

bayindirh• 1 year ago

There's also $300/mo AI ULTRA membership. It's interesting. Google One memberships even can't detail what "extra features" I can have, because it possibly changes every hour or so.

Keyframe2• 1 year ago

There's also $300/mo AI ULTRA membership

Not if you're in EU though. Even though I have zero or less AI use so far, I tinker with it. I'm more than happy to pay $200+tax for Max 20x. I'd be happy to pay same-ish for Gemini Pro.. if I knew how and where to have Gemini CLI like I do with Claude code. I have Google One. WHERE DO I SIGN UP, HOW DO I PAY AND USE IT GOOGLE? Only thing I have managed so far is through openrouter via API and credits which would amount to thousands a month if I were to use it as such, which I won't do.

What I do now is occasionally I go to AI Studio and use it for free.

SecretDreams• 1 year ago

Maybe their products team is also just run by Gemini, and it's changing its mind every day?

I also just got the email for Gemini ultra and I couldn't even figure out what was being offered compared to pro outside of 30tb storage vs 2tb storage!

gavinray• 1 year ago

I actually had this exact same question when I read the docs, made an issue about it:

https://github.com/google-gemini/gemini-cli/issues/1427

chrisandchris• 1 year ago

I find it very interesting that the only way to reach Google support .... seems to be to file an issue in a competitors product (besides using X or whatever).

3abiton• 1 year ago

And they say our scale up is siloed. Leave it to google to show' em.

nojito5• 1 year ago

You don't get API keys for that subscription because it's a flat monthly cost.

iandanforth2• 1 year ago

That's not a given, Anthropic recently added Claude CLI access to their $20/m "Pro" plan removing the need for a separate API key.

behnamoh6• 1 year ago

Actually, that's the reason a lot of startups and solo developers prefer non-Google solutions, even though the quality of Gemini 2.5 Pro is insanely high. The Google Cloud Dashboard is a mess, and they haven't fixed it in years. They have Vertex that is supposed to host some of their models, but I don't understand what's the difference between that and their own cloud. And then you have two different APIs depending on the level of your project: This is literally the opposite of what we would expect from an AI provider where you start small and regardless of the scale of your project, you do not face obstacles. So essentially, Google has built an API solution that does not scale because as soon as your project gets bigger, you have to switch from the Google AI Studio API to the Vertex API. And I find it ridiculous because their OpenAI compatible API does not work all the time. And a lot of tools that rely on that actually don't work.

Google's AI offerings that should be simplified/consolidated:

- Jules vs Gemini CLI?

- Vertex API (requires a Google Cloud Account) vs Google AI Studio API

Also, since Vertex depends on Google Cloud, projects get more complicated because you have to modify these in your app [1]:

``` # Replace the `GOOGLE_CLOUD_PROJECT` and `GOOGLE_CLOUD_LOCATION` values # with appropriate values for your project. export GOOGLE_CLOUD_PROJECT=GOOGLE_CLOUD_PROJECT export GOOGLE_CLOUD_LOCATION=global export GOOGLE_GENAI_USE_VERTEXAI=True ```

[1] https://cloud.google.com/vertex-ai/generative-ai/docs/start/...

tarvaina• 1 year ago

It took me a while but I think the difference between Vertex and Gemini APIs is that Vertex is meant for existing GCP users and Gemini API for everyone else. If you are already using GCP then Vertex API works like everything else there. If you are not, then Gemini API is much easier. But they really should spell it out, currently it's really confusing.

Also they should make it clearer which SDKs, documents, pricing, SLAs etc apply to each. I still get confused when I google up some detail and end up reading the wrong document.

coredog64• 1 year ago

At least a bunch of people got promotions for demonstrating scope via the release of a top-level AI product.

cperry27• 1 year ago

@sachinag is afk but wanted me to flag that he's on point for fixing the Cloud Dashboard - it's WIP!

irthomasthomas3• 1 year ago

I just use gemini-pro via openrouter API. No painful clicking around on the cloud to find the billing history.

__MatrixMan__3• 1 year ago

Anthropic is the same. Unless it has changed within the last few months, you can subscribe to Claude but if you want to use Claude Code it'll come out of your "API usage" bucket which is billed separately than the subscription.

Some jerk has learned that we prefer CLI things and has come to the conclusion that we should therefore pay extra for them.

Workaround is to use their GUI with some MCPs but I dislike it because window navigation is just clunky compared to terminal multiplexer navigation.

gnur• 1 year ago

This has changed actually, since this month you can use claude code if you have a cloud pro subscription.

trostaft• 1 year ago

AFAIK, Claude code operates on your subscription, no? That's what this support page says

https://support.anthropic.com/en/articles/11145838-using-cla...

Could have changed recently. I'm not a user so I can't verify.

kissgyorgy2• 1 year ago

This is simply not true. All personal paid packages include Claude Code now.

unshavedyak3• 1 year ago

In addition to others mentioning subscriptions being better in Claude Code, i wanted to compare the two so i tried to find a Claude Max equivalent license... i have no clue how. In their blog post they mention `Gemini Code Assist Standard or Enterprise license` but they don't even link to it.. lol.

Some googling lands me to a guide: https://cloud.google.com/gemini/docs/discover/set-up-gemini#...

I stopped there because i don't want to signup i just wanted to review, but i don't have an admin panel or etc.

It feels insane to me that there's a readme on how to give them money. Claude's Max purchase was just as easy as Pro, fwiw.

Workaccount26• 1 year ago

I think it is pretty clear that these $20/subs are loss leaders, and really only meant to get regular people to really start leaning on LLMs. Once they are hooked, we will see what the actual price of using so much compute is. I would imagine right now they are pricing their APIs either at cost or slightly below.

carefulfungi• 1 year ago

This is down voted I guess because the circumstances have changed - but boy is it still confusing. All these platforms have chat subscriptions, api pay-as-you-go, CLI subscriptions like "claude code" ... built-in offers via Github enterprise or Google Workspace enterprise ...

It's a frigg'n mess. Everyone at our little startup has spent time trying to understand what the actual offerings are; what the current set of entitlements are for different products; and what API keys might be tied to what entitlements.

I'm with __MatrixMan__ -- it's super confusing and needs some serious improvements in clarity.

HarHarVeryFunny2• 1 year ago

Isn't that a bit like saying that gasoline should be sold as a fixed price subscription rather than a usage based scheme where long distance truckers pay more than someone driving < 100 miles per week?

A ChatBot is more like a fixed-price buffet where usage is ultimately human limited (even if the modest eaters are still subsidizing the hogs). An agentic system is going to consume resources in much more variable manner, depending on how it is being used.

Some jerk has learned that we prefer CLI things and has come to the conclusion that we should therefore pay extra for them

Obviously these companies want you to increase the amount of their product you consume, but it seems odd to call that a jerk move! FWIW, Anthropic's stated motivation for Claude Code (which Gemini is now copying) was be agnostic to your choice of development tools since CLI access is pretty much ubiquitous, even inside IDEs. Whether it's the CLI-based design, the underlying model, or the specifics of what Claude Code is capable of, they seem to have got something right, and apparently usage internal to Anthropic skyrocketed just based on word of mouth.

GardenLetter272• 1 year ago

Google is fumbling the bag so badly with the pricing.

Gemini 2.5 Pro is the best model I've used (even better than o3 IMO) and yet there's no simple Claude/Cursor like subscription to just get full access.

Nevermind Enterprise users too, where OpenAI has it locked up.

llm_nerd3• 1 year ago

I wouldn't dream of thinking anyone has anything "locked up". Certainly not OpenAI which increasingly seems to be on an uphill battle against competitors (including Microsoft who even though they're a partner, are also a competitor) who have other inroads.

Not sure what you mean by "full access", as none of the providers offer unrestricted usage. Pro gets you 2.5 Pro with usage limits. Ultra gets you higher limits + deep think (edit: accidentally put research when I meant think where it spends more resources on an answer) + much more Veo 3 usage. And of course you can use the API usage-billed model.

bachmeier5• 1 year ago

Google is fumbling the bag so badly with the pricing.

In certain areas, perhaps, but Google Workspace at $14/month not only gives you Gemini Pro, but 2 TB of storage, full privacy, email with a custom domain, and whatever else. College students get the AI pro plan for free. I recently looked over all the options for folks like me and my family. Google is obviously the right choice, and it's not particularly close.

bcrosby95• 1 year ago

They're 'fumbling' because these models are extremely expensive to run. It's also why there's so many products and so much confusion across the whole industry.

Xmd5a2• 1 year ago

Gemini 2.5 Pro is the best model

It's the second time I read this in this thread. May I ask why you think this is the case? And in which domains? I am very satisfied with 2.5 pro when it comes to philosophical/literary analysis, probably because of the super long context I can fill with whole books, and wanted to try Claude Code for the same purpose, but with folders, summaries, etc to make up for the shorter context length.

Davidzheng4• 1 year ago

Just go on AI Studio?

ur-whale4• 1 year ago

Delightful!

You clearly have never had the "pleasure" to work with a Google product manager.

Especially the kind that were hired in the last 15-ish years.

This type of situation is absolutely typical, and probably one of the more benign thing among the general blight they typically inflict on Google's product offering.

The cartesian product of pricing options X models is an effing nightmare to navigate.

bachmeier5• 1 year ago

I had a conversation with Copilot about Copilot offerings. Here's what they told me:

If I Could Talk to Satya...

I'd say:

“Hey Satya, love the Copilots—but maybe we need a Copilot for Copilots to help people figure out which one they need!”

Then I had them print out a table of Copilot plans:

- Microsoft Copilot Free - Github Copilot Free - Github Copilot Pro - Github Copilot Pro+ - Microsoft Copilot Pro (can only be purchased for personal accounts) - Microsoft 365 Copilot (can't be used with personal accounts and can only be purchased by an organization)

boston_clone2• 1 year ago

I'd really like to hear your own personal perspective on the topic instead of a regurgitation of an LLM.

diegof792• 1 year ago

Google suffers from Microsoft's issues: it has products for almost everything, but its confusing product messaging dilutes all the good things it does.

I like Gemini 2.5 Pro, too, and recently, I tried different AI products (including the Gemini Pro plan) because I wanted a good AI chat assistant for everyday use. But I also wanted to reduce my spending and have fewer subscriptions.

The Gemini Pro subscription is included with Google One, which is very convenient if you use Google Drive. But I already have an iCloud subscription tightly integrated with iOS, so switching to Drive and losing access to other iCloud functionality (like passwords) wasn’t in my plans.

Then there is the Gemini chat UI, which is light years behind the OpenAI ChatGPT client for macOS.

NotebookLM is good at summarizing documents, but the experience isn’t integrated with the Gemini chat, so it’s like constantly switching between Google products without a good integrated experience.

The result is that I end up paying a subscription to Raycast AI because the chat app is very well integrated with other Raycast functions, and I can try out models. I don’t get the latest model immediately, but it has an integrated experience with my workflow.

My point in this long description is that by being spread across many products, Google is losing on the UX side compared to OpenAI (for general tasks) or Anthropic (for coding). In just a few months, Google tried to catch up with v0 (Google Stitch), GH Copilot/Cursor (with that half-baked VSCode plugin), and now Claude Code. But all the attempts look like side-projects that will be killed soon.

Fluorescence3• 1 year ago

The Gemini Pro subscription is included with Google One

It's not in Basic, Standard or Premium.

It's in a new tier called "Google AI Pro" which I think is worth inclusion in your catalogue of product confusion.

Oh wait, there's even more tiers that for some reason can't be paid for annually. Weird... why not? "Google AI Ultra" and some others just called Premium again but now include AI. 9 tiers, 5 called Premium, 2 with AI in the name but 6 that include Gemini. What a mess.

krferriter• 1 year ago

I subscribed to Google One through the Google Photos iOS app because I wanted photos I took on my iPhone to be backed up to Google. When I switched to Android and went into Google One to increase my storage capacity in my Google account, I found that it was literally impossible, because the subscription was tied to my iCloud account. I even got on a line with Google Support about it and they told me yeah it's not even possible on their side to disconnect my Google One subscription from Apple. I had to wait for the iCloud subscription to Google One to end, and then I was able to go into Google One and increase my storage capacity.

pjmlp• 1 year ago

It is quite interesting how many big corporations eventually grow up to the same mess.

Yizahi• 1 year ago

I had no idea I had a Gemini Pro sub included in One sub all this time. Thanks for the tip :)

Google is really bad at effective advertising.

Imustaskforhelp• 1 year ago

There is a vscode extension that can basically be an agent but use gemini from the website which is cool.

But I found it to a little bit clunky and I guess I like the ui of google, I mean, the point is to get the point across. If you really hate the gemini ui, I am pretty sure that there is stylus extension which can beautify it or change the styles to your looking.

I guess I am an android user but still I understand your icloud subscription but if you're only choice as to why to not switch to google is passwords (but maybe you can mention more?), then for passwords, please try bitwarden, I found it to be really delightful.

UrineSqueegee2• 1 year ago

losing access to other iCloud functionality (like passwords) wasn’t in my plans.

you can export and import the passwords and you can sync your photos to google photos

guestbest• 1 year ago

You’d think with all this AI tooling they'd be able to organize better, but I think that the AI Age will be a very messy one with messaging and content

tmaly• 1 year ago

I was just trying to figure out if I get anything as a pro user. Thank you, you answered my question.

This is very confusing how they post about this on X, you would think you get additional usage. Messaging is very confusing.

upcoming-sesame4• 1 year ago

I think Pro is for regular folks, not specifically for programmers.

I also have a pro subscription and wish I could get an API key with that with generous quota as well but pro is just for "consumers" using Gemini app I guess

fhinkel17• 1 year ago

That's valuable feedback and we're taking it to heart.

tom_m2• 1 year ago

They are horrendous about their product strategy and positioning.

pnarotam001• 1 year ago

Google needs to package their products better. It's always been the case with them. GCP is confusing because you can't easily find their services.

ulnarkressty• 1 year ago

The other day I was trying to figure out a cost estimate for Gemini 2.5 Pro but they don't have this in their pricing calculator - it just isn't there. Only endless tables with various model sizes and options which aren't explained.

artdigital2• 1 year ago

How can I subscribe to this? From the comments here it sounds like it's NOT bundled into the Gemini Pro subscription, but into "Gemini Code Assist Standard".

I can't find any way to upgrade to a paid plan, is this even possible for individuals, or is it just "free or enterprise"?

/Edit: Okay I went through the Gemini docs. I found that in Google Cloud you can enable Gemini Code Assist Standard and Enterprise for the account

- Standard is $19.00/mo

- Enterprise is $45.00/mo

Difference between the 2 editions: https://cloud.google.com/products/gemini/pricing

/Edit2: Found the actual management screen: https://codeassist.google.com/overview

powerapple• 1 year ago

Google does not care about users at all, ZERO, it is full of people are so fond of technology, they try something and soon move on to something else. I am using Google Cloud for work, and previously Google Mail , Google Map, and YouTube user, zero good experience.

2muchcoffeeman• 1 year ago

I was having such a good time using Gemini 2.5 Pro I forgot that Google is generally incapable of sustaining a new product. I’ll try not to get used too attached to it.

rhodysurf2• 1 year ago

I neeeed this google login method in sst's opencode now haha

ZeroCool2u5• 1 year ago

Ugh, I really wish this had been written in Go or Rust. Just something that produces a single binary executable and doesn't require you to install a runtime like Node.

iainmerrick5• 1 year ago

Looks like you could make a standalone executable with Bun and/or Deno:

https://bun.sh/docs/bundler/executables

https://docs.deno.com/runtime/reference/cli/compile/

Note, I haven't checked that this actually works, although if it's straightforward Node code without any weird extensions it should work in Bun at least. I'd be curious to see how the exe size compares to Go and Rust!

ZeroCool2u5• 1 year ago

Yeah, this just seems like a pain in the ass that could've been easily avoided.

JimDabell3• 1 year ago

I was going to say the same thing, but they couldn’t resist turning the project into a mess of build scripts that hop around all over the place manually executing node.

buildfocus2• 1 year ago

You can also do this natively with Node, since v18: https://nodejs.org/api/single-executable-applications.html#s...

tln• 1 year ago

A Bun "hello world" is 58Mb

Claude also requires npm, FWIW.

fhinkel17• 1 year ago

Ask Gemini CLI to re-write itself in your preferred language

ZeroCool2u5• 1 year ago

Unironically, not a bad idea.

shitpostbot• 1 year ago

Let's not pretend this stuff actually works

AJ0072• 1 year ago

Contest between Claude Code and Gemini CLI, who rewrites it faster/cheaper/better?

frollogaston11• 1 year ago

I have already been asking the Gemini CLI questions about itself.

geodel• 1 year ago

My thoughts exactly. Neither Rust not Go, not even C/C++ which I could accept if there were some native OS dependencies. Maybe this is a hint on who could be its main audience.

ur-whale4• 1 year ago

Maybe this is a hint on who could be its main audience.

Or a hint about the background of the folks who built the tool.

wiseowise7• 1 year ago

Definitely not neckbeards who care about such trivia.

qsort• 1 year ago

Projects like this have to update frequently, having a mechanism like npm or pip or whatever to automatically handle that is probably easier. It's not like the program is doing heavy lifting anyway, unless you're committing outright programming felonies there shouldn't be any issues on modern hardware.

It's the only argument I can think of, something like Go would be goated for this use case in principle.

ZeroCool2u5• 1 year ago

I feel like Cargo or Go Modules can absolutely do the same thing as the mess of build scripts they have in this repo perfectly well and arguably better.

koakuma-chan4• 1 year ago

If you use Node.js your program is automatically too slow for a CLI, no matter what it actually does.

masklinn2• 1 year ago

having a mechanism like npm or pip or whatever to automatically handle that is probably easier

Re-running `cargo install <crate>` will do that. Or install `cargo-update`, then you can bulk update everything.

And it works hella better than using pip in a global python install (you really want pipx/uvx if you're installing python utilities globally).

IIRC you can install Go stuff with `go install`, dunno if you can update via that tho.

mpeg• 1 year ago

You'd think that, but a globally installed npm package is annoying to update, as you have to do it manually and I very rarely need to update other npm global packages so at least personally I always forget to do it.

frollogaston11• 1 year ago

I don't think that's the main reason. Just installed this and peaked in node_nodules. There are a lot of random deps, probably for the various local capabilities, and it was probably easier to find those libs in the Node ecosystem than elsewhere.

Also, react-reconciler caught my eye. Apparently that's a dependency of ink, which lets you write text-based UIs in React.

That and opentelemetry, whatever the heck that is

js2• 1 year ago

Build updating into the tool. e.g.

  uv self update
  yt-dlp --update

etc.

i_love_retros2• 1 year ago

This isn't about quality products, it's about being able to say you have a CLI tool because the other ai companies have one

clbrmbr• 1 year ago

Fast following is a reasonable strategy. Anthropic provided the existence proof. It’s an immensely useful form factor for AI.

behnamoh6• 1 year ago

This isn't about quality products, it's about being able to say you have a CLI tool because the other ai companies have one

Anthropic's Claude Code is also installed using npm/npx.

rs1862• 1 year ago

Eh, I can't see how your comment is relevant ti the parent thread. Creating a CLI in Go is barely more complicated than JS. Rust, probably, but people aren't asking for that.

frollogaston11• 1 year ago

Writing it in Golang or Rust doesn't really make it better

buildfocus2• 1 year ago

Node can also produce a single binary executable: https://nodejs.org/api/single-executable-applications.html

ur-whale4• 1 year ago

and doesn't require you to install a runtime like Node.

My exact same reaction when I read the install notes.

Even python would have been better.

Having to install that Javascript cancer on my laptop just to be able to try this, is a huge no.

jart4• 1 year ago

See gemmafile which gives you an airgapped version of gemini (which google calls gemma) that runs locally in a single file without any dependencies.

https://huggingface.co/jartine/gemma-2-27b-it-llamafile

quotemstr2• 1 year ago

Language choice is orthogonal to distribution strategy. You can make single-file builds of JavaScript (or Python or anything) programs! It's just a matter of packaging, and there are packaging solutions for both Bun and Node. Don't blame the technology for people choosing not to use it.

frollogaston11• 1 year ago

Why would you want a single-file build anyway, to make it easier to move around on disk? There are reasons for the dep filetree.

Btw, the largest deps in this are React and Open Telemetry.

corysama• 1 year ago

Meanwhile, https://analyticsindiamag.com/global-tech/openai-is-ditching...

I really don't mind either way. My extremely limited experience with Node indicates they have installation, packaging and isolation polished very well.

frollogaston11• 1 year ago

Node and Rust both did packaging well, I think Golang too. It's a disaster in Python.

8n4vidtmkvmk5• 1 year ago

Bun can compile to a single executable. Not sure if node has the same feature. Point is it's very doable with JS.

truth_seeker• 1 year ago

You mean something like this:

https://www.npmjs.com/package/pkg

or perhaps this one:

https://www.npmjs.com/package/nexe

qudat• 1 year ago

Why would someone use this over aider?

adamtaylor_13• 1 year ago

Disclaimar: I haven't used aider in probably a year. I found Aider to require much more understanding to use properly. Claude code _just works_, more or less out of the box. Assuming the Gemini team took cues from CC—I'm guessing it's more user-friendly than Aider.

Again, I haven't used aider in a while so perhaps that's not the case.

bananapub2• 1 year ago

Claude Code and OpenAI Codex and presumably this are much much more aggressive about generating work for themselves than Aider is.

For complicated changes Aider is much more likely to stop and need help, whereas Claude Code will just go and go and end up with something.

Whether that's worth the different economic model is up to you and your style and what you're working on.

fhinkel17• 1 year ago

I played around with it to automate GitHub tasks for me (tagging and sorting PRs and stuff). Sometimes it needs a little push to use the API instead of web search, but then it even installs the right tools (like gh) for you. https://youtu.be/LP1FtpIEan4

cperry27• 1 year ago

Hi - I work on this. Uptake is a steep curve right now, spare a thought for the TPUs today.

Appreciate all the takes so far, the team is reading this thread for feedback. Feel free to pile on with bugs or feature requests we'll all be reading.

elashri• 1 year ago

Hi, Thanks for this work.

currently it seems these are the CLI tools available. Is it possible to extend or actually disable some of these tools (for various reasons)?

Available Gemini CLI tools:

    - ReadFolder
    - ReadFile
    - SearchText
    - FindFiles
    - Edit
    - WriteFile
    - WebFetch
    - ReadManyFiles
    - Shell
    - Save Memory
    - GoogleSearch

_ryanjsalva3• 1 year ago

I also work on the product. You can extend the tools with MCP. https://github.com/google-gemini/gemini-cli/blob/main/docs/t...

cperry27• 1 year ago

I had to ask Gemini CLI to remind myself ;) but you can add this into settings.json:

{ "excludeTools": ["run_shell_command", "write_file"] }

but if you ask Gemini CLI to do this it'll guide you!

bdmorgan3• 1 year ago

I also work on the product :-)

You can also extend with the Extensions feature - https://github.com/google-gemini/gemini-cli/blob/main/docs/e...

SafeDusk• 1 year ago

Pretty close to what I discovered is essential in https://github.com/aperoc/toolkami, 7 tools will cover majority of the use cases.

ebiester2• 1 year ago

So, as a member of an organization who pays for google workspace with gemini, I get the message `GOOGLE_CLOUD_PROJECT environment variable not found. Add that to your .env and try again, no reload needed!`

At the very least, we need better documentation on how to get that environment variable, as we are not on GCP and this is not immediately obvious how to do so. At the worst, it means that your users paying for gemini don't have access to this where your general google users do.

cperry27• 1 year ago

https://github.com/google-gemini/gemini-cli/blob/main/docs/c...

thimabi15• 1 year ago

I believe Workspace users have to pay a separate subscription to use the Gemini CLI, the so-called “Gemini for Google Cloud”, which starts at an additional 19 dollars per month [^1]. If that’s really the case, it’s very disappointing to me. I expected access to Gemini CLI to be included in the normal Workspace subscription.

[^1]: https://console.cloud.google.com/marketplace/product/google/...

fooey• 1 year ago

workspace accounts always seems like an unsupported mess at google, which is a very strange strategy

carraes• 1 year ago

it would be cool to work with my google ai pro sub

cperry27• 1 year ago

working on it

nojito5• 1 year ago

How often did you use gemini-cli to build on gemini-cli?

_ryanjsalva3• 1 year ago

We started using Gemini CLI to build itself after about week two. If I had to guess, I'd say better than 80% of the code was written with Gemini CLI. Honestly, once we started using the CLI, we started experimenting a lot more and building waaaaay faster.

bdmorgan3• 1 year ago

100% of the time

mkagenius• 1 year ago

Hi - I integrated Apple Container on M1 to run[1] the code generated by Gemini CLI. It works great!

1. CodeRunner - https://github.com/BandarLabs/coderunner/tree/main?tab=readm...

cperry27• 1 year ago

<3 amazing

javier123454321• 1 year ago

one piece of feedback. Please do neovim on top of vim or have a way to customize the editor beyond your list.

newnimzo• 1 year ago

someone has already sent out a PR for this! https://github.com/google-gemini/gemini-cli/pull/1448

bsenftner• 1 year ago

Thank you for your work on this. I spent the afternoon yesterday trying to convert an algorithm written in ruby (which I do not know) to vanilla JavaScript. It was a comedy of failing nonsense as I tried to get gpt-4.1 to help, and it just led me down pointless rabbit holes. I installed Gemini CLI out of curiosity, pointed it at the Ruby project, and it did the conversion from a single request, total time from "think I'll try this" to it working was 5 minutes. Impressed.

cperry27• 1 year ago

<3 love to hear it!

dtech• 1 year ago

I did something very similar last week with Claude Code, which also had good results. Good to know Gemini CLI is on the same level of power.

willy_k2• 1 year ago

FWIW the fair comparison for Gemini 2.5 Pro would be o4-mini. That being said I’ve also found that Gemini is way better at getting it right on the first or second try and responding to corrections.

streb-lo• 1 year ago

Is there a reason all workspace accounts need a project ID? We pay for gemini pro for our workspace accounts but we don't use GCP or have a project ID otherwise.

thimabi15• 1 year ago

The reason is that billing is separate, via the paid tier of the API. Just a few minutes ago, I was able to test Gemini CLI using a Workspace account after setting up a project in the free tier of the API. However, that seems to have been a bug on their end, because I now get 403 errors (Forbidden) with that configuration. The remaining options are either to set up billing for the API or use a non-Workspace Google account.

cperry27• 1 year ago

the short answer is b/c one of our dependencies requires it and hasn't resolved it.

danavar• 1 year ago

Is there a way to instantly, quickly prompt it in the terminal, without loading the full UI? Just to get a short response without filling the terminal page.

like to just get a short response - for simple things like "what's a nm and grep command to find this symbol in these 3 folders". I use gemini alot for this type of thing already

Or would that have to be a custom prompt I write?

peterldowns• 1 year ago

I use `mods` for this https://github.com/charmbracelet/mods

other people use simon willison's `llm` tool https://github.com/simonw/llm

Both allow you to switch between models, send short prompts from a CLI, optionally attach some context. I prefer mods because it's an easier install and I never need to worry about Python envs and other insanity.

cperry27• 1 year ago

-p is your friend

hiAndrewQuinn2• 1 year ago

gemini --prompt "Hello"

irthomasthomas3• 1 year ago

If you uv install llm Then grab my shelllm scripts github.com/irthomasthomas/shelllm and source them in your terminal then you can use premade prompt functions like shelp "what's a nm and grep command to find this symbol in these 3 folders" -m gemini-pro

There's also wrappers that place the command directly in your terminal prompt if you run shelp-c

conception2• 1 year ago

Google Gemini Google Gemini Ultra AI Studio Vertex AI Notebook LLM Jules

All different products doing the sameish thing. I don’t know where to send users to do anything. They are all licensed differently. Bonkers town.

sagarm2• 1 year ago

Is it really that confusing? Gemini is the equivalent of ChatGPT; AI Studio is for advanced users that want to control e.g. temperature; Vertex AI is the GCP integrated API; Notebook LLM is basically personal RAG; and Jules is a developer agent.

Many of these are not even remotely similar.

GenerWork• 1 year ago

I'm just a hobbyist, but I keep getting the error "The code change produced by Gemini cannot be automatically applied. You can manually apply the change or ask Gemini to try again". I assume this is because the service is being slammed?

Edit: I should mention that I'm accessing this through Gemini Code Assist, so this may be something out of your wheelhouse.

cperry27• 1 year ago

odd, haven't seen that one - you might file an issue https://github.com/google-gemini/gemini-cli/issues

I don't think that's capacity, you should see error codes.

taupi• 1 year ago

Right now authentication doesn't work if you're working on a remote machine and try to authenticate with Google, FYI. You need an alternate auth flow that gives the user a link and lets them paste a key in (this is how Claude Code does it).

cperry27• 1 year ago

correct, sorry, known issue

ciwchris• 1 year ago

Using the Gemini CLI the first thing I tried to do was "Create GEMINI.md files to customize your interactions with Gemini." The command ran for about a minute before receiving a too many requests error.

You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.

Discouraging

fhinkel17• 1 year ago

Super weird! I've been using it the last week, and never hit the quota limit for free users. We're having some capacity issues right now, but that should not affect the quota. Would love it if you can try tomorrow or so again!

hiAndrewQuinn2• 1 year ago

Feature request! :)

I'm a Gemini Pro subscriber and I would love to be able to use my web-based chat resource limits with, or in addition to, what is offered here. I have plenty of scripts that are essentially "Weave together a complex prompt I can send to Gemini Flash to instantly get the answer I'm looking for and xclip it to my clipboard", and this would finally let me close the last step in that scripts.

Love what I'm seeing so far!

cperry27• 1 year ago

working on it!

imjonse• 1 year ago

Hi. It is unclear from the README whether the free limits apply also when there's an API key found in the environment - not explicitly set for this tool - and there is no login requirement.

cperry27• 1 year ago

if you explicitly select the sign in with google you'll get the free tier - it won't use your API key.

Freedom2• 1 year ago

Pointed it at a project directory and asked it to find and fix an intentionally placed bug without referencing any filenames. It seemed to struggle finding any file or constructing a context about the project unless specifically asked. FWIW, Claude Code tries to build an 'understanding' of the codebase when given the same prompt. For example, it struggled when I asked to "fix the modal logic" but nothing was specifically called a modal.

Is the recommendation to specifically ask "analyze the codebase" here?

yomismoaqui• 1 year ago

I have been evaluating other tools like Amp (from Sourcegraph) and when trying Gemini Cli on VsCode I found some things to improve:

- On a new chat I have to re-approve things like executing "go mod tidy", "git", write files... I need to create a new chat for each feature, (maybe an option to clear the current chat on VsCode would work)

- I have found some problems with adding some new endpoint on an example Go REST server I was trying it on, it just deleted existing endpoints on the file. Same with tests, it deleted existing tests when asking to add a test. For comparison I didn't find these problems when evaluating Amp (uses Claude 4)

Overall it works well and hope you continue with polishing it, good job!!

cperry27• 1 year ago

thank you kind stranger!

sandGorgon• 1 year ago

i have a Google AI Pro subscription - what kind of credits/usage/allowance do i get towards gemini cli ?

cperry27• 1 year ago

not connected yet

kingsleyopara2• 1 year ago

Thanks so much for this! I’d really appreciate a more consumer oriented subscription offering, similar to Claude Max, that combines Gemini CLI (with IP compliance) and the Gemini app (extra points for API access too!).

cperry27• 1 year ago

working on it

upcoming-sesame4• 1 year ago

this seems to be the number one topic in this thread

Xmd5a2• 1 year ago

Hey the interface on YouTube loads super slowly for me. The change appeared a few months ago. I'm not talking about the video streams, but the ajax loading of the UI. Whether it's opening a new youtube tab or navigating between videos within youtube, it takes forever. Chrome/Safari -> same deal, 30 seconds delays is what I observe. My macbook pro is 10 years old, the problem doesn't appear on more recent hardware, but still youtube shouldn't be the slowest website to load on my machine. I can load spotify.com just fine in about 5 seconds.

jadbox• 1 year ago

Does it have LSP (language server) support? How should I think of this as different from Aider?

nprateem6• 1 year ago

Please, for the love of God, stop your models always answering with essays or littering code with tutorial style comments. Almost every task devolves into "now get rid of the comments". It seems impossible to prevent this.

And thinking is stupid. "Show me how to generate a random number in python"... 15s later you get an answer.

msgodel• 1 year ago

They have to do that, it's how they think. If they were trained not to do that they'd produce lower quality code.

mpalmer3• 1 year ago

Take some time to understand how the technology works, and how you can configure it yourself when it comes to thinking budget. None of these problems sound familiar to me as a frequent user of LLMs.

kridsdale35• 1 year ago

Congrats on your success. May you all be promoted and may your microkitchens be stocked.

cperry27• 1 year ago

thank you kind stranger! may gemini only return you perfection on the first response.

atlgator• 1 year ago

Thank you for supporting Windows natively.

owebmaster2• 1 year ago

Why does this CLI don't support google's A2A?

jskherman• 1 year ago

Hello, thanks for the work for finally having an analog to Claude Code.

A natural question to ask is, if in the near future, can Google One "Google AI Pro" subscribers have higher limits than what is offered for free users?

manmal• 1 year ago

The thing that had me close it and go back to Claude Code immediately was how often Gemini CLI failed to make file changes. This doesn’t instill confidence at all, with it retrying edits multiple times - who knows what it ends up editing, then.

CC has this issue too, but way less often, and second shot almost always works.

bloppe7• 1 year ago

Requests for documentation clarification:

- Here [1] it says "Project settings override user settings." How does gemini determine if we're in a project? Does it look for a `.gemini` folder in the current working directory as well as every parent directory up to the root? Would Gemini be able to read the contents of a subfolder of the CWD if the subfolder contains a different `.gemini` folder?

- I don't see documentation for the `selectedAuthType` field in the documentation for settings.json. Mine says `oauth-personal`. I could've sworn I signed in with my Google Workspace account. Does `oauth-personal` apply to Workspace accounts?

And a feature request: it would be nice to add a restriction in the settings.json file forcing anybody who uses gemini in that project to sign in to a Workspace account in a specific org (or use a specific project, I guess).

[1] https://github.com/google-gemini/gemini-cli/blob/main/docs/c...

blitzar• 1 year ago

F for the TPUs.

ripped_britches• 1 year ago

Hi cperry - do you have an idea of when Gemini 2.5 live API will make GA? I saw a new GA label was created last week.

redhale2• 1 year ago

Is LiteLLM integration on the roadmap? I get that part of the motivation to build this is to funnel usage to Google models, but it would be nice to be able to go through OpenRouter and/or use local models too. Open source is nice, but open model configuration would be nicer :)

bravura• 1 year ago

Thanks for building this! The tool shows a lot of promise. Coming from Claude Code, the core functionality feels solid - just needs some permission refinements to match enterprise use cases. This is based upon quickly skimming the current code.

High ROI feature requests:

• Pattern-based permissions - Bash(git:) to allow git but not rm, Write(logs/.txt) for path scoping

• CLI permission flags - --allowedTools "Read,Bash(npm test)" --deniedTools "Write" for session overrides

• Allow/deny precedence rules - explicit deny should override general allow (security principle)

• Config file hierarchy - system → user → project precedence for policy enforcement

Medium ROI improvements:

• Command argument filtering - whitelist git commit but not git --exec-path=/bin/sh

• Multiple config formats - support both simple arrays and structured permission objects

• Runtime permission diagnostics - gemini permissions list to debug what's actually enabled

• Environment variable injection - top-level env config for OTEL endpoints, API keys, etc.

The permission engine is really the key piece - once you can express "allow X but not Y within X", it unlocks most advanced use cases. Keep up the great work!

Phlogistique• 1 year ago

On the one hand, yes this has obviously high immediate value; on the other hand, I can't help but feel like you are giving access to multiple tools that can be used for arbitrary code execution anyway (i.e. running tests, installing dependencies, or running any linter that has a plugin system...), so blacklisting `git --exec-path=/bin/sh` for example is... Misguided? You would have a better time containing the agent in an environment without internet access?

cperry27• 1 year ago

feedback is a gift - thank you!!

akrauss3• 1 year ago

One thing I'd really like to see in coding agents is this: As an architect, I want to formally define module boundaries in my software, in order to have AI agents adhere to and profit from my modular architecture.

Even with 1M context, for large projects, it makes sense to define boundaries These will typically be present in some form, but they are not available precisely to the coding agent. Imagine there was a simple YAML format where I could specify modules and where they can be found in the source tree, and the APIs of other modules it interacts with. Then it would be trivial to turn this into a context that would very often fit into 1M tokens. When an agent decides something needs to be done in the context of a specific module, it could then create a new context window containing exactly that module, effetively turning a large codebase into a small codebase, for which Gemini is extraordinarily effective.

akrauss3• 1 year ago

There is one feature in Claude Code which is often overlooked and I haven't seen it in any of the other agentic tools: There is a tool called "sub-agent", which creates a fresh context windows in which the model can independently work on a clearly defined sub-task. This effectively turns Claude Code from a single-agent model to a hierarchical multi-agent model (I am not sure if the hierarchy goes to depths >2).

I wonder if it is a concious decision not to include this (I imagine it opens a lot of possibilities of going crazy, but it also seems to be the source of a great amount of Claud Code's power). I would very much like to play with this if it appears in gemini-cli

Next step would be the possibility to define custom prompts, toolsets and contexts for specific re-occuring tasks, and these appearing as tools to the main agent. Example for such a thing: create_new_page. The prompt could describe the steps one needs to create the page. Then the main agent could simply delegate this as a well-defined task, without cluttering its own context with the operational details.

indigodaddy5• 1 year ago

Would Gemini non-interactive mode be a stop gap if they don't have sub-agent equivalent yet?

https://github.com/google-gemini/gemini-cli/blob/main/docs/c...

ramirond• 1 year ago

"sub-agent" sounds incredible! All tools should implement that.

cperry27• 1 year ago

conscious decision not to include it mostly to cut a release we could ship to land yesterday ;)

various forms of this are being discussed, this commentary is helpful thanks!

leguy• 1 year ago

Could Gemini CLI be used for PR reviews? For example, would you expect that asking Gemini to compare two git commit hashes and analyze the code changes for potential bugs/conciseness/adhesion to the idiomatic project style, etc. to work well?

Edit: I tried it. The setup was a breeze. I fed the CLI two git commit IDs and some light prompting on what to look for. It gave a reasonable response. I'll try on a real PR shortly.

desireco42• 1 year ago

Overall it looks and feels good. I gave it a problem, like, what was it... to update website pages to use single layout, it wasn't trivial but it wasn't that hard. It burned through 7M tokens and like 20-25mins didn't accomplished it. I stopped it because I didn't want to waste more resources.

I think with better prompting on my end, as I have good experience with Gemini, this will be awesome. You probably could tweak a lot on your end as well, don't let it get stuck in cycles.

tealpod• 1 year ago

Two weeks back I made a simple wrapper CLI app for GEMINI https://geni.dev.

I strongly believe in AI cli apps for devs, congrats.

meetpateltech• 1 year ago

Key highlights from blog post and GitHub repo:

- Open-source (Apache 2.0, same as OpenAI Codex)

- 1M token context window

- Free tier: 60 requests per minute and 1,000 requests per day (requires Google account authentication)

- Higher limits via Gemini API or Vertex AI

- Google Search grounding support

- Plugin and script support (MCP servers)

- Gemini.md file for memory instruction

- VS Code integration (Gemini Code Assist)

edg50003• 1 year ago

This sounds like it's practically unlimited and free? A bit like GMail. It is comes to close to Claude Code they may overtake them and dominate the market simply with their tried and true "give it away for free forever with practially no limits" strategy. I pay 20$ for Claude Code, which is nothing in relation to the value it provides, but obviously a 0$ price point will get you a bigger piece of the Pie. Big Oof for Anthropic. Kinda lame that they can just copy a new and innovative company and undersell at a price point of 0$. Not that I'm cheering for Anthropic either; I'd much rather own my sillicon.

i_love_retros2• 1 year ago

Boring. Any non llm news?

ape4• 1 year ago

In the screenshot it's asked about Gemini CLI and it says its going to search the web and read the README.md - what ever did we do before AI /s

titusblair• 1 year ago

Nice work excited to use it!

andrewstuart• 1 year ago

I really wish these AI companies would STOP innovating until they work out how to let us “download all files” on the chat page.

We are now three years into the AI revolution and they are still forcing us to copy and paste and click click crazy to get the damn files out.

STOP innovating. STOP the features.

Form a team of 500 of your best developers. Allocate a year and a billion dollar budget.

Get all those Ai super scientists into the job.

See if you can work out “download all files”. A problem on the scale of AGI or Dark Matter, but one day google or OpenAI will crack the problem.

nojito5• 1 year ago

This edits the files directly. Using a chat hasn't been an optimal workflow for a while now.

raincole4• 1 year ago

What does this even mean lol. "Download all files"...?

Workaccount26• 1 year ago

It seems you are still using the web interface.

When you hop over to platforms that use the API, the files get written/edited in situ. No copy/pasting. No hunting for where to insert edited code.

Trust me it's a total game changer to switch. I spent so much time copy/pasting before moving over.

indigodaddy5• 1 year ago

https://github.com/robertpiosik/CodeWebChat

matltc• 1 year ago

Sweet, I love Claude and was raring to try out their CLI that dropped a few days ago, but don't have a sub. This looks to be free

lazarie• 1 year ago

"Failed to login. Ensure your Google account is not a Workspace account."

Is your vision with Gemini CLI to be geared only towards non-commercial users? I have had a workspace account since GSuite and have been constantly punished for it by Google offerings all I wanted was gmail with a custom domain and I've lost all my youtube data, all my fitbit data, I cant select different versions of some of your subscriptions (seemingly completely random across your services from a end-user perspective), and now as a Workspace account I cant use Gemini CLI for my work, which is software development. This approach strikes me as actively hostile towards your loyal paying users...

GlebOt• 1 year ago

Have you checked the https://github.com/google-gemini/gemini-cli/blob/main/docs/c... ? It has a section for workspace accounts.

Aeolun3• 1 year ago

It shouldn’t be that hard. Logically it should just be, signin and go.

raincole4• 1 year ago

It seems that you need to set up an env variable called GOOGLE_CLOUD_PROJECT https://github.com/google-gemini/gemini-cli/issues/1434

... and other stuff.

LouisvilleGeek• 1 year ago

The barrier to use this project is maddening. I went through all of the setup instructions and getting the workspace error for a personal gmail account.

Googlers, we should not have to do all of this setup and prep work for a single account. Enterprise I get, but for a single user? This is insufferable.

zxspectrum19822• 1 year ago

Same here.

jsnell2• 1 year ago

What's up with printing lame jokes every few seconds? The last thing I want from a tool like this is my eye to be drawn to the window all the time as if something had changed and needs my action. (Having a spinner is fine, having changing variable length text isn't.)

asadm10• 1 year ago

can disable it from accessibility setting. it does show model thinking instead of joke when that’s available.

jsnell2• 1 year ago

Thanks, but where are those accessibility settings? /help shows nothing related to settings other than auth and theme, there's no related flags, and there's a ~/.gemini/settings.json that contains just the auth type.

No mention of accessibility in https://github.com/google-gemini/gemini-cli/blob/0915bf7d677... either

phillipcarter2• 1 year ago

An aside, but with Claude Code and now Gemini instrumenting operations with OpenTelemetry by default, this is very cool.

giuscri• 1 year ago

can you elaborate?

phillipcarter2• 1 year ago

Aside from being an opentelemetry maintainer who likes seeing these things ...

... it's a big deal because the only way to really make sense of using LLMs in any context is to have good observability data you can analyze later. And so the traces that they emit here will show all the CLI invocations, inputs/outputs and context for each chat turn in a session, and you can look for patterns that exhibit good or bad stuff.

Jayakumark• 1 year ago

Whether any CLI interactions are used to train the model or no ?

imiric• 1 year ago

Ha. It would be naive to think that a CLI tool from an adtech giant won't exploit as much data as it can collect.

thimabi15• 1 year ago

You raise an interesting topic. Right now, when we think about privacy in the AI space, most of the discussion hinges on using our data for training purposes or not. That being said, I figure it won’t be long before AI companies use the data they collect to personalize ads as well.

rbren• 1 year ago

If you're looking for a fully open source, LLM-agnostic alternative to Claude Code and Gemini CLI, check out OpenHands: https://docs.all-hands.dev/usage/how-to/cli-mode

joelthelion• 1 year ago

Or aider. In any case, while top llms will likely remain proprietary for some time, there is no reason for these tools to be closed source or tied to a particular llm vendor.

spiffytech• 1 year ago

I've had a good experience with https://kilocode.ai

It integrates with VS Code, which suits my workflow better. And buying credits through them (at cost) means I can use any model I want without juggling top-ups across several different billing profiles.

lostmsu3• 1 year ago

Can it use Claude sub or Gemini free tier the same way Gemini CLI does?

lostmsu3• 1 year ago

How does its CLI mode compare to Claude Code and Gemini CLI?

rhodysurf2• 1 year ago

opencode.ai

mekpro• 1 year ago

Just refactored 1000 lines of Claude Code generated to 500 lines with Gemini Pro 2.5 ! Very impressed by the overall agentic experience and model performance.

barbazoo• 1 year ago

To use Gemini CLI free-of-charge, simply login with a personal Google account to get a free Gemini Code Assist license. That free license gets you access to Gemini 2.5 Pro and its massive 1 million token context window. To ensure you rarely, if ever, hit a limit during this preview, we offer the industry’s largest allowance: 60 model requests per minute and 1,000 requests per day at no charge.

If it sounds too good to be true, it probably is. What’s the catch? How/why is this free?

raincole4• 1 year ago

Because Google is rich and they'd like to get you hooked. Just like how ChatGPT has a free tier.

Also they can throttle the service whenever they feel it's too costly.

leumon• 1 year ago

My guess: So that they can get more training data to improve their models which will eventually be subscription only.

jabroni_salad• 1 year ago

They recently discontinued the main Gemini free tier which offered similar limits. I would say expect this to disappear when it hits GA or if it gets a lot of targeted abuse.

dawnofdusk3• 1 year ago

https://en.wikipedia.org/wiki/First-mover_advantage

lostmsu3• 1 year ago

I just tried it - it does not last long.

jmkni• 1 year ago

Also I have Gemini Pro, what am I paying for?

iaresee2• 1 year ago

Whoa. Who at Google thought providing this as an example of how to test your API key was a good idea?

https://imgur.com/ZIZkLU7

This is shown at the top of the screen in https://aistudio.google.com/apikey as the suggested quick start for testing your API key out.

Not a great look. I let our GCloud TAM know. But still.

asadm10• 1 year ago

What's wrong here?

iaresee2• 1 year ago

Don't put your API keys as parameters in your URL. Great way to have them land in server logs, your shell history, etc. You're trusting no one with decryption capabilities is doing logging and inspection correctly, which you shouldn't.

nickysielicki• 1 year ago

it's wrapped in TLS, is ok.

Mond_• 1 year ago

Oh hey, afaik all of this LLM traffic goes through my service!

Set up not too long ago, and afaik pretty load-bearing for this. Feels great, just don’t ask me any product-level questions. I’m not part of the Gemini CLI team, so I’ll try to keep my mouth shut.

Not going to lie, I’m pretty anxious this will fall over as traffic keeps climbing up and up.

asadm10• 1 year ago

do you mean the genai endpoints?

paradite• 1 year ago

I chanced up some endpoint that mentioned "makersuite" that seems to be what's behind genai.

It's interesting to see how one team / product gets rebranded to another in Google.

cperry27• 1 year ago

thank you for your service. I too have been anxious all day :)

acedTrex• 1 year ago

Everyone writing the same thing now lol, its plainly obvious this is the workflow best suited to llms

albertzeyer• 1 year ago

The API can be used both via your normal Google account, or via API key?

Because it says in the README:

Authenticate: When prompted, sign in with your personal Google account. This will grant you up to 60 model requests per minute and 1,000 model requests per day using Gemini 2.5 Pro.
For advanced use or increased limits: If you need to use a specific model or require a higher request capacity, you can use an API key: ...

When I have the Google AI Pro subscription in my Google account, and I use the personal Google account for authentication here, will I also have more requests per day then?

I'm currently wondering what makes more sense for me (not for CLI in particular, but for Gemini in general): To use the Google AI Pro subscription, or to use an API key. But I would also want to use the API maybe at some point. I thought the API requires an API key, but here it seems also the normal Google account can be used?

bdmorgan3• 1 year ago

It's firmly on the radar - we will have a great answer for this soon.

rtaylorgarlock3• 1 year ago

I spent 8k tokens after giving the interface 'cd ../<other-dir>', resulting in Gemini explaining that it can't see the other dir outside of current scope but with recommendation ls files in that dir. Which then reminded me of my core belief that we will always have to be above these tools in order to understand and execute. I wonder if/when I'll be wrong.

0x4572• 1 year ago

Well, I'm happy about sandboxing. Idk what is your issue? 8k tokens?

revskill• 1 year ago

Nice, at least i could get rid of the broken Warp CLI which prevents offline usage with their automatic cloud ai feature enabled.

zxspectrum19822• 1 year ago

Does Gemini CLI require API access?

frereubu• 1 year ago

I have access to Gemini through Workspace, but despite spending quite a while trying to find out how, I cannot figure out how to use that in Copilot. All I seem to be able to find is information on the personal account or enterprise tiers, neither of which I have.

wohoef• 1 year ago

A few days ago I tested Claude Code by completely vibe coding a simple stock tracker web app in streamlit python. It worked incredibly well, until it didn't. Seems like there is a critical project size where it just can't fix bugs anymore. Just tried this with Gemini CLI and the critical project size it works well for seems to be quite a bit bigger. Where claude code started to get lost, I simply told Gemini CLI to "Analyze the codebase and fix all bugs". And after telling it to fix a few more bugs, the application simply works.

We really are living in the future

AJ0072• 1 year ago

Current best practice for Claude Code is to have heavy lifting done by Gemini Pro 2.5 or o3/o3pro. There are ways to do this pretty seamlessly now because of MCP support (see Repo Prompt as an example.) Sometimes you can also just use Claude but it requires iterations of planning, integration while logging everything, then repeat.

I haven't looked at this Gemini CLI thing yet, but if its open source it seems like any model can be plugged in here?

I can see a pathway where LLMs are commodities. Every big tech company right now both wants their LLM to be the winner and the others to die, but they also really, really would prefer a commodity world to one where a competitor is the winner.

If the future use looks more like CLI agents, I'm not sure how some fancy UI wrapper is going to result in a winner take all. OpenAI is winning right now with user count by pure brand name with ChatGPT, but ChatGPT clearly is an inferior UI for real work.

sysmax• 1 year ago

I think, there are different niches. AI works extremely well for Web prototyping because a lot of that work is superficial. Back in the 90s we had Delphi where you could make GUI applications with a few clicks as opposed to writing tons of things by hand. The only reason we don't have that for Web is the decentralized nature of it: every framework vendor has their own vision and their own plan for future updates, so a lot of the work is figuring out how to marry the latest version of component X with the specific version of component Y because it is required by component Z. LLMs can do that in a breeze.

But in many other niches (say embedded), the workflow is different. You add a feature, you get weird readings. You start modelling in your head, how the timing would work, doing some combination of tracing and breakpoints to narrow down your hypotheses, then try them out, and figure out what works the best. I can't see the CLI agents do that kind of work. Depends too much on the hunch.

Sort of like autonomous driving: most highway driving is extremely repetitive and easy to automate, so it got automated. But going on a mountain road in heavy rain, while using your judgment to back off when other drivers start doing dangerous stuff, is still purely up to humans.

sagarpatil• 1 year ago

You might want to give this a try: https://github.com/opencode-ai/opencode

jswny• 1 year ago

Which MCPs allow you to for the heavy lifting with external models in Claude Code?

dawnofdusk3• 1 year ago

I feel like you get more mileage out of prompt engineering and being specific... not sure if "fix all the bugs" is an effective real-world use case.

tvshtr3• 1 year ago

Yeah, and it's variable, can happen at 250k, 500k or later. When you interrogate it; usually the issue comes to it being laser focused or stuck on one specific issue, and it's very hard to turn it around. For the lack of the better comparison it feels like the AI is on a spectrum...

ugh123• 1 year ago

Claude seems to have trouble with extracting code snippets to add to the context as the session gets longer and longer. I've seen it get stuck in a loop simply trying to use sed/rg/etc to get just a few lines out of a file and eventually give up.

TechDebtDevin• 1 year ago

Yeah but this collapses under any real complexity and there is likely an extreme amount of redundant code and would probably be twice as memory efficient if you just wrote it yourself.

Im actually interested to see if we see a rise in demand for DRAM that is greater than usual because more software is vibe coded than being not, or some form of vibe coding.

crazylogger• 1 year ago

Ask the AI to document each module in a 100-line markdown. These should be very high level, don't contain any detail, but just include pointers to relevant files for AI to find out by itself. With a doc as the starting point, AI will have context to work on any module.

If the module just can't be documented in this way in under 100 lines, it's a good time to refactor. Chances are if Claude's context window is not enough to work with a particular module, a human dev can't either. It's all about pointing your LLM precisely at the context that matters.

agotterer2• 1 year ago

I wonder how much of this had to do with the context window size? Gemini’s window is 5x larger than Cladue’s.

I’ve been using Claude for a side project for the past few weeks and I find that we really get into a groove planning or debugging something and then by the time we are ready to implement, we’ve run out of context window space. Despite my best efforts to write good /compact instructions, when it’s ready to roll again some of the nuance is lost and the implementation suffers.

I’m looking forward to testing if that’s solved by the larger Gemini context window.

macNchz• 1 year ago

I definitely think the bigger context window helps. The code quality quite visibly drops across all models I've used as the context fills up, well before the hard limit. The editor tooling also makes a difference—Claude Code pollutes its own context window with miscellaneous file accesses and tool calls as it tries to figure out what to do. Even if it's more manual effort to manage the files that are in-context with Aider, I find the results to be much more consistent when I'm able to micromanage the context.

Approaching the context window limit in Claude Code, having it start to make more and worse mistakes, then seeing it try to compact the context and keep going, is a major "if you find yourself in a hole, stop digging" situation.

data-ottawa2• 1 year ago

Does /compact help with this? I ran out of context with claude code for the first time today, so looking for any tips.

I'm trying to get better at the /resume and memories to try and get more value out of the tool.

seunosewa5• 1 year ago

I've found that I can quickly get a new AI session up to speed by adding critical context that it's missing. In my largest codebase it's usually a couple of critical functions.once they have the key context, they can do the rest. This of course doesn't work when you can't view their thinking process and interrupt it to supply them with the context that they are missing. Opacity doesn't work unless the agent does the right thing every time.

tobyhinloopen5• 1 year ago

At some point, LLMs just get distracted and you might be better off throwing it away and restarting hah.

KETpXDDzR• 1 year ago

IIRC Claude Code indexes your code via a vector DB. That's why it's good for larger codebases. I'm not aware that GeminiCLI has this, too.

bufo• 1 year ago

Grateful that this one supports Windows out of the box.

ruffrey• 1 year ago

Thanks, Google. A bit of feedback - integration with `gcloud` CLI auth would have been appreciated.

koakuma-chan4• 1 year ago

It doesn't work. It just gives me 429 after a minute.

fhinkel17• 1 year ago

We're working on it! The response has been incredible, so if you don't want to wait you can also get started with an API key from: https://aistudio.google.com/app/apikey Apologies!

nostrebored• 1 year ago

is there anywhere to track the qos degradation? would love to use this for a feature we're shipping today just to try it out, but consistently get 429's on Gemini or Vertex. Checking quotas shows that neither are close to limits, which makes me think it's the DSP being infra constrained??

ptone4• 1 year ago

Try it out with 0.1.5 https://github.com/google-gemini/gemini-cli/discussions/2064

solomatov• 1 year ago

I couldn't find any mentions of whether they train their models on your source code. May be someone was able to?

dawnofdusk3• 1 year ago

Yes they do. Scroll to bottom of Github readme

This project leverages the Gemini APIs to provide AI capabilities. For details on the terms of service governing the Gemini API, please refer to the terms for the access mechanism you are using:

Click Gemini API, scroll

When you use Unpaid Services, including, for example, Google AI Studio and the unpaid quota on Gemini API, Google uses the content you submit to the Services and any generated responses to provide, improve, and develop Google products and services and machine learning technologies, including Google's enterprise features, products, and services, consistent with our Privacy Policy.
To help with quality and improve our products, human reviewers may read, annotate, and process your API input and output. Google takes steps to protect your privacy as part of this process. This includes disconnecting this data from your Google Account, API key, and Cloud project before reviewers see or annotate it. Do not submit sensitive, confidential, or personal information to the Unpaid Services.

jddj• 1 year ago

There must be thousands of keys in those logs.

Workaccount26• 1 year ago

If you use for free: Yes

If you pay for API: No

jonnycoder• 1 year ago

The plugin is getting bad reviews this morning. It doesn't work for me on latest Pycharm.

mil225• 1 year ago

Does anyone know what Google's policy on retention and training use will be when using the free version by signing in with a personal Google account? Like many others, I don't want my proprietary codebase stored permanently on Google servers or used to train their models.

At the bottom of README.md, they state:

"This project leverages the Gemini APIs to provide AI capabilities. For details on the terms of service governing the Gemini API, please refer to the terms for the access mechanism you are using:

* Gemini API key

* Gemini Code Assist

* Vertex AI"

The Gemini API terms state: "for Unpaid Services, all content and responses is retained, subject to human review, and used for training".

The Gemini Code Assist terms trifurcate for individuals, Standard / Enterprise, and Cloud Code (presumably not relevant).

* For individuals: "When you use Gemini Code Assist for individuals, Google collects your prompts, related code, generated output, code edits, related feature usage information, and your feedback to provide, improve, and develop Google products and services and machine learning technologies."

* For Standard and Enterprise: "To help protect the privacy of your data, Gemini Code Assist Standard and Enterprise conform to Google's privacy commitment with generative AI technologies. This commitment includes items such as the following: Google doesn't use your data to train our models without your permission."

The Vertex AI terms state "Google will not use Customer Data to train or fine-tune any AI/ML models without Customer's prior permission or instruction."

What a confusing array of offerings and terms! I am left without certainty as to the answer to my original question. When using the free version by signing in with a personal Google account, which doesn't require a Gemini API key and isn't Gemini Code Assist or Vertex AI, it's not clear which access mechanism I am using or which terms apply.

It's also disappointing "Google's privacy commitment with generative AI technologies" which promises that "Google doesn't use your data to train our models without your permission" doesn't seem to apply to individuals.

Oras• 1 year ago

Appreciate how easy it is to report a bug! I like these commands.

A bit gutted by the `make sure it is not a workspace account`. What's wrong with Google prioritising free accounts vs paid accounts? This is not the first time they have done it when announcing Gemini, too.

ranuzz• 1 year ago

It says in the pricing page (https://ai.google.dev/gemini-api/docs/pricing) that free usage can be used to improve products, may be for them it makes sense to beta release to free users first https://ai.google.dev/gemini-api/terms

cperry27• 1 year ago

you can use it with workspace, just need to pay. I'm told this is just a temporary limitation we're looking to resolve.

iddan• 1 year ago

This is awesome! We recently started using Xander (https://xander.bot). We've found it's even better to assign PMs to Xander on Linear comments and get a PR. Then, the PM can validate the implementation in a preview environment, and engineers (or another AI) can review the code.

stpedgwdgfhgdd2• 1 year ago

Another JS implementation…

I do not get it why they don’t pick Go or Rust so i get a binary.

llm_nerd3• 1 year ago

Given that there's another comment complaining about this being in node...

This perfectly demonstrates the benefit of the nodejs platform. Trivial to install and use. Almost no dependency issues (just "> some years old version of nodejs"). Immediately works effortlessly.

I've never developed anything on node, but I have it installed because so many hugely valuable tools use it. It has always been absolutely effortless and just all benefit.

And what a shift from most Google projects that are usually a mammoth mountain of fragile dependencies.

(uv kind of brings this to python via uvx)

ekunazanu2• 1 year ago

I have nothing against npm, but having a single binary would've been a lot more trivial and convenient. Codex is heading in that direction, and I hope others do too.

llm_nerd3• 1 year ago

I don't see how it can get more trivial and convenient than

    npm install -g @google/gemini-cli

then having a gemini command line app. I mean, I guess if you don't have node installed, but it's so prevalent in so many tools it's a reasonable thing to require.

Piping a downloaded shell script to bash seems pretty much the same effort.

ivanjermakov• 1 year ago

Gemini, convert my disk from MBR to GPT

jeffbee• 1 year ago

Unironically. The new Ubuntu installer has such a piss-poor UI that I, a 33-year Linux user, could not figure out how to get it to partition my disk, and searching the web turned up nothing other than spam and video spam, until I ordered Gemini to give me gparted commands that achieved what I wanted and made the Ubuntu UI unblock itself.

Keyframe2• 1 year ago

Hmm, with Claude code at $200+tax, this seems to be alternative which comes out at free or $299+tax a YEAR if I need more which is great. I found that buried at developers.google.com

Gemini Pro and Claude play off of each other really well.

Just started playing with Gemini CLI and one thing I miss immediately from Claude code is being able to write and interject as the AI does its work. Sometimes I interject by just saying stop, it stops and waits for more context or input or ai add something I forgot and it picks it up..

jilles• 1 year ago

Anyone else think it's interesting all these CLIs are written in TypeScript? I'd expect Google to use Go.

Aeolun3• 1 year ago

How am I supposed to use this when actually working on a cli? The sign in doesn’t display s link I can open. Presumably it’s trying and failing to open firefox?

oc1• 1 year ago

one would expect that google engineers would know that a cli tool should have code flow auth instead of forcing you to open the default browser on the same machine which defeats the purpose of a cli for many use cases

ipsum210• 1 year ago

If you use this, all of your code data will be sent to Google. From their terms:

https://developers.google.com/gemini-code-assist/resources/p...

When you use Gemini Code Assist for individuals, Google collects your prompts, related code, generated output, code edits, related feature usage information, and your feedback to provide, improve, and develop Google products and services and machine learning technologies.

To help with quality and improve our products (such as generative machine-learning models), human reviewers may read, annotate, and process the data collected above. We take steps to protect your privacy as part of this process. This includes disconnecting the data from your Google Account before reviewers see or annotate it, and storing those disconnected copies for up to 18 months. Please don't submit confidential information or any data you wouldn't want a reviewer to see or Google to use to improve our products, services, and machine-learning technologies.

jart4• 1 year ago

Mozilla and Google provide an alternative called gemmafile which gives you an airgapped version of Gemini (which Google calls Gemma) that runs locally in a single file without any dependencies. https://huggingface.co/jartine/gemma-2-27b-it-llamafile It's been deployed into production by 32% of organizations: https://www.wiz.io/reports/the-state-of-ai-in-the-cloud-2025

nicce2• 1 year ago

That is just Gemma model. Most people seek capabilities equivalent for Gemini 2.5 Pro if they want to do any kind of coding.

ipsum210• 1 year ago

There's nothing wrong with promoting your own projects, but its a little weird that you don't disclose that you're the creator.

rudedogg2• 1 year ago

Insane to me there isn’t even an asterisk in the blog post about this. The data collection is so over the top I don’t think users suspect it because it’s just absurd. For instance Gemini Pro chats are trained on too.

If this is legal, it shouldn’t be.

mattzito3• 1 year ago

It's a lot more nuanced than that. If you use the free edition of Code Assist, your data can be used UNLESS you opt out, which is at the bottom of the support article you link to:

"If you don't want this data used to improve Google's machine learning models, you can opt out by following the steps in Set up Gemini Code Assist for individuals."

and then the link: https://developers.google.com/gemini-code-assist/docs/set-up...

If you pay for code assist, no data is used to improve. If you use a Gemini API key on a pay as you go account instead, it doesn't get used to improve. It's just if you're using a non-paid, consumer account and you didn't opt out.

That seems different than what you described.

ipsum210• 1 year ago

Sorry, that's not correct. Did you check out the link? It doesn't describe the CLI, only the IDE.

"You can find the Gemini Code Assist for individuals privacy notice and settings in two ways:

- VS Code - IntelliJ "

foob• 1 year ago

your data can be used UNLESS you opt out

It's even more nuanced than that.

Google recently testified in court that they still train on user data after users opt out from training [1]. The loophole is that the opt-out only applies to one organization within Google, but other organizations are still free to train on the data. They may or may not have cleaned up their act given that they're under active investigation, but their recent actions haven't exactly earned them the benefit of the doubt on this topic.

[1] https://www.business-standard.com/technology/tech-news/googl...

aflukasz• 1 year ago

It's a lot more nuanced than that. If you use the free edition of Code Assist, your data can be used UNLESS you opt out,

Well... you are sending your data to a remote location that is not yours.

andrepd• 1 year ago

Yes, I'm right about to trust Google to do what they pinky swear.

EDIT: Lmao, case in point, two sibling comments pointing out that Google does indeed do this anyway via some loophole; also they can just retain the data and change the policy unilaterally in the future.

If you want privacy do it local with Free software.

Workaccount26• 1 year ago

This is just for free use (individuals), for standard and enterprise they don't use the data.

Which pretty much means if you are using it for free, they are using your data.

I don't see what is alarming about this, everyone else has either the same policy or no free usage. Hell the surprising this is that they still let free users opt-out...

thimabi15• 1 year ago

everyone else has either the same policy or no free usage

That’s not true. ChatGPT, even in the free tier, allows users to opt out of data sharing.

nojito5• 1 year ago

If you use this, all of your code data will be sent to Google.

Not if you pay for it.

reaperducer2• 1 year ago

If you use this, all of your code data will be sent to Google.

Not if you pay for it.

Today.

In six months, a "Terms of Service Update" e-mail will go out to an address that is not monitored by anyone.

FiberBundle• 1 year ago

Do you honestly believe that the opt-out by Anthropic and Cursor means your code won't be used for training their models? Seems likely that they would rather just risk taking a massive fine for potentially solving software development than to let some competitor try it instead.

rudedogg2• 1 year ago

Yes.

The resulting class-action lawsuit would bankrupt the company, along with the reputation damage, and fines.

olejorgenb2• 1 year ago

For API users, we automatically delete inputs and outputs on our backend within 30 days of receipt or generation, except when you and we have agreed otherwise (e.g. zero data retention agreement), if we need to retain them for longer to enforce our Usage Policy (UP), or comply with the law.

If this is due to compliance with law I wonder how they can make the zero-data-retention agreement work... The companies I've seen have this have not mention that they themself retain the data...

mil225• 1 year ago

They really need to provide some clarity on the terms around data retention and training, for users who access Gemini CLI free via sign-in to a personal Google account. It's not clear whether the Gemini Code Assist terms are relevant, or indeed which of the three sets of terms they link at the bottom of the README.md apply here.

fhinkel17• 1 year ago

Agree! We're working on it!

mil225• 1 year ago

There is some information on this buried in configuration.md under "Usage Statistics". They claim:

*What we DON'T collect:*

- *Personally Identifiable Information (PII):* We do not collect any personal information, such as your name, email address, or API keys.

- *Prompt and Response Content:* We do not log the content of your prompts or the responses from the Gemini model.

- *File Content:* We do not log the content of any files that are read or written by the CLI.

https://github.com/google-gemini/gemini-cli/blob/0915bf7d677...

jdironman• 1 year ago

I wonder what the legal difference between "collect" and "log" is.

ipsum210• 1 year ago

This is useful, and directly contradicts the terms and conditions for Gemini CLI (edit: if you use the personal account, then its governed under the Code Assist T&C). I wonder which one is true?

naiv• 1 year ago

Who cares, software has no value anymore.

crat3r• 1 year ago

Sarcasm? Weird statement if not.

I still have yet to replace a single application with an LLM, except for (ironically?) Google search.

I still use all the same applications as part of my dev work/stack as I did in the early 2020's. The only difference is occasionally using an LLM baked into to one of them but the reality is I don't do that much.

FL4102• 1 year ago

To be honest this is by far the most frustrating part of the Gemini ecosystem, to me. I think 2.5 pro is probably the best model out there right now, and I'd love to use it for real work, but their privacy policies are so fucking confusing and disjointed that I just assume there is no privacy whatsoever. And that's with the expensive Pro Plus Ultra MegaMax Extreme Gold plan I'm on.

I hope this is something they're working on making clearer.

dmbche• 1 year ago

If I'm being cynical, it's easy to either say "we use it" or "we don't touch it" but they'd lose everyone that cares about this question if they just said "we use it" - most beneficial position is to keep it as murky as possible.

If I were you I'd assume they're using all of it for everything forever and act accordingly.

ipsum210• 1 year ago

In my own experience, 2.5 Pro 03-26 was by far the best LLM model at the time.

The newer models are quantized and distilled (I confirmed this with someone who works on the team), and are a significantly worse experience. I prefer OpenAI O3 and o4-mini models to Gemini 2.5 Pro for general knowledge tasks, and Sonnet 4 for coding.

UncleOxidant2• 1 year ago

For coding in my experience Claude Sonnet/Opus 4.0 is hands down better than Gemini 2.5. pro. I just end up fighting with Claude a lot less than I do with Gemini. I had Gemini start a project that involved creating a recursive descent parser for a language in C. It was full of segfaults. I'd ask Gemini to fix them and it would end up breaking something else and then we'd get into a loop. Finally I had Claude Sonnet 4.0 take a look at the code that Gemini had created. It fixed the segfaults in short order and was off adding new features - even anticipating features that I'd be asking for.

BryanLegend• 1 year ago

Some of my code is so bad I'm sure it will damage their models!

nprateem6• 1 year ago

I'm not that bothered. Most of it came from Google or anthropic anyway

fnl• 1 year ago

How does that compare to Claude Code? How protected are you when using CC?

Buttons840• 1 year ago

Is this significantly different than what we agree to when we put code on GitHub?

fhinkel17• 1 year ago

Hey all, This is a really great discussion, and you've raised some important points. We realize the privacy policies for the Gemini CLI were confusing depending on how you log in, and we appreciate you calling that out.

To clear everything up, we've put together a single doc that breaks down the Terms of Service and data policies for each account type, including an FAQ that covers the questions from this thread.

Here’s the link: https://github.com/google-gemini/gemini-cli/blob/main/docs/t...

Thanks again for pushing for clarity on this!

HenriNext• 1 year ago

Thanks, one more clarification please. The heading of point #3 seems to mention Google Workspace: "3. Login with Google (for Workspace or Licensed Code Assist users)". But the text content only talks about Code Assist: "For users of Standard or Enterprise edition of Gemini Code Assist" ... Could you clarify whether point #3 applies with login via Google Workspace Business accounts?

ipsum210• 1 year ago

I think you did a good job CYA on this, but what people were really looking for was a way to opt-out of Google collecting code, similar to the opt-out process for the IDE is available.

FL4102• 1 year ago

I appreciate the effort here, but I’m still confused. Is my $250/mo “ultra” plan considered personal and still something you train on?

yuuluuu• 1 year ago

Thanks for clarification!

kiitos• 1 year ago

Is there any way for a user using the "Login with Google ... for individuals" auth method (I guess auth method 1) -- to opt-out of, and prevent, their input prompts, and output responses, from being used as training data?

From an initial parse of your linked tos-privacy.md doc, it seems like the answer is "no" -- but that seems bonkers to me, so I hope I'm misreading or misunderstanding something!

predkambrij• 1 year ago

seems to be straight "yes" with no opt-out. https://github.com/google-gemini/gemini-cli/blob/main/docs/t...

fastball• 1 year ago

Kinda a tragedy of the commons situation. Everyone wants to use these tools that must be trained on more and more code to get better, but nobody wants it to be trained on their code. Bit silly imo.

learner0073• 1 year ago

good to know, I won't be using this. Curious, do you know if OpenAI codex and Claude also do the same? I was under the impression that they don't share code.

jonbaer2• 1 year ago

-y, --yolo Automatically accept all actions (aka YOLO mode, see https://www.youtube.com/watch?v=xvFZjo5PgG0 for more details)? [boolean] [default: false]

cperry27• 1 year ago

:) thank you for discovering our extensive documentation on this valued feature

jonbaer2• 1 year ago

What about ... https://github.com/google-gemini/gemini-cli/blob/swarm-mode-...

alpb• 1 year ago

Are there any LLMs that offer ZSH plugins that integrate with command history, previous command outputs, system clipboard etc to assist writing the next command? Stuff like gemini/copilot CLI don't feel particularly useful to me. I'm not gonna type "?? print last 30 lines of this file"

rapatel0• 1 year ago

not zsh but you might want to try warp 2.0

logicchains• 1 year ago

That's giving a lot away for free! When I was using Gemini 2.5 Pro intensively for automated work and regularly hitting the 1000 requests per day limit, it could easily cost $50+ per day with a large context. I imagine after a couple months they'll probably limit the free offering to a cheaper model.

WaltPurvis• 1 year ago

They said in the announcement that these high usage limits are only while it's in the preview stage, so they may (or may not) be reduced later. I think (it's not clear, at least to me) that the CLI may be using cheaper models already, at times. It says if you want the freedom to choose your model for every request, you can buy one of the paid subscriptions. I interpret that to mean that some of your requests may go to Gemini 2.5 Pro, but not all of them, and it decides. Am I extremely wrong about that?

beboplifa• 1 year ago

Wow, this is next-level. I can't believe this is free. This blows cline out of the water!

lordofgibbons• 1 year ago

How does this compare to OpenCode and OAI's Codex? Those two are also free, they work with any LLM.

https://github.com/opencode-ai/opencode

kissgyorgy2• 1 year ago

Codex is just terrible

joelm• 1 year ago

Been using Claude Code (4 Opus) fairly successfully in a large Rust codebase, but sometimes frustrated by it with complex tasks. Tried Gemini CLI today (easy to get working, which was nice) and it was pretty much a failure. It did a notably worse job than Claude at having the Rust code modifications compile successfully.

However, Gemini at one point output what will probably be the highlight of my day:

"I have made a complete mess of the code. I will now revert all changes I have made to the codebase and start over."

What great self-awareness and willingness to scrap the work! :)

ZeroCool2u5• 1 year ago

Personally my theory is that Gemini benefits from being able to train on Googles massive internal code base and because Rust has been very low on uptake internally at Google, especially since they have some really nice C++ tooling, Gemini is comparatively bad at Rust.

dilap• 1 year ago

That's interesting. I've tried Gemini 2.5 Pro from time to time because of the rave reviews I've seen, on C# + Unity code, and I've always been disappointed (compared to ChatGPT o3 and o4-high-mini and even Grok). This would support that theory.

danielbln3• 1 year ago

Interesting, Gemini must be a monster when it comes to Go code then. I gotta try it for that

thimabi15• 1 year ago

Personally my theory is that Gemini benefits from being able to train on Googles massive internal code base

But does Google actually train its models on its internal codebase? Considering that there’s always the risk of the models leaking proprietary information and security architecture details, I hardly believe they would run that risk.

leoh7• 1 year ago

Personally my theory is that Gemini benefits from being able to train on Googles massive internal code base and because Rust has been very low on uptake internally at Google, especially since they have some really nice C++ tooling, Gemini is comparatively bad at Rust.

Were they to train it on their C++ codebase, it would not be effective on account of the fact that they don't use boost or cmake or any major stuff that C++ in the wider world use. It would also suggest that the user make use of all kinds of non-available C++ libraries. So no, they are not training on their own C++ corpus nor would it be particularly useful.

data-ottawa2• 1 year ago

Tangental, but I worry that LLMs will cause a great stagnation in programming language evolution, and possibly a bunch of tech.

I've tried using a few new languages and the LLMs would all swap the code for syntactically similar languages, even after telling them to read the doc pages.

Whether that's for better or worse I don't know, but it does feel like new languages are genuinely solving hard problems as their raison d'etre.

joshvm• 1 year ago

Gemini has some fun failure modes. It gets "frustrated" when changes it makes doesn't work, and replies with oddly human phrases like "Well, that was unexpected" and then happily declares that (I see the issue!) "the final tests will pass" when it's going down a blind alley. It's extremely overconfident by default and much more exclamatory without changing the system prompt. Maybe in training it was taught/figured out that manifesting produces better results?

jjice• 1 year ago

It also gets really down on itself, which is pretty funny (and a little scary). Aside from the number of people who've posted online about it wanting to uninstall itself after being filled with shame, I had it get confused on some Node module resolution stuff yesterday and it told me it was deeply sorry for wasting my time and that I didn't deserve to have such a useless assistant.

Out of curiosity, I told it that I was proud of it for trying and it had a burst of energy again and tried a few more (failing) solution, before going back to it's shameful state.

Then I just took care of the issue myself.

raincole4• 1 year ago

So far I've found Gemini CLI is very good at explaining what existing code does.

I can't say much about writing new code though.

fpgaminer• 1 year ago

Claude will do the same start over if things get too bad. At least I've seen it when its edits went haywire and trashed everything.

eknkc• 1 year ago

Same here. Tried to implement a new feature on one of our apps to test it. It completely screwed things up. Used undefined functions and stuff. After a couple of iterations of error reporting and fixing I gave up.

Claude did it fine but I was not happy with the code. What Gemini came up with was much better but it could not tie things together at the end.

taberiand• 1 year ago

Sounds like you can use gemini to create the initial code, then have claude review and finalise what gemini comes up with

skerit• 1 year ago

I tried it too, it was so bad. I got the same "revert" behaviour after only 15 minutes.

noisy_boy• 1 year ago

I asked it to do a comparatively pedestrian task: write a script to show top 5 google searches.

First it did the search itself and then added "echo" for each of them - cute

Then it tried to use pytrends which didn't go anywhere

Then it tried some other paid service which also didn't go anywhere

Then it tried some other stuff which also didn't go anywhere

Finally it gave up and declared failure.

It will probably be useful as it can do the modify/run loop itself with all the power of Gemini but so far, underwhelming.

fcoury• 1 year ago

This was also my exact experience. I was pretty excited because I usually use Gemini Pro 2.5 when Claude Code gets stuck by pasting the whole code and asking questions and it was able to get me out of a few pickles a couple of times.

Unfortunately the CLI version wasn't able to create coherent code or fix some issues I had in my Rust codebase as well.

Here's hope that it eventually becomes great.

incomingpain• 1 year ago

Giving this a try, I'm rather astounded how effective my tests have gone.

That's a ton of free limit. This has been immensely more successful than void ide.

dmd• 1 year ago

Well, color me not impressed. On my very first few tries, out of 10 short-ish (no more than 300 lines) python scripts I asked it to clean up and refactor, 4 of them it mangled to not even run any more, because of syntax (mostly quoting) errors and mis-indenting. Claude has never done that.

atonse• 1 year ago

This is all very cool, but I hate to be the "look at the shiny lights" guy...

How did they do that pretty "GEMINI" gradient in the terminal? is that a thing we can do nowadays? It doesn't seem to be some blocky gradient where each character is a different color. It's a true gradient.

(yes I'm aware this is likely a total clone of claude code, but still am curious about the gradient)

krat0sprakhar• 1 year ago

Look at the package.json inside the CLI folder https://www.npmjs.com/package/ink-gradient

ars• 1 year ago

Maybe this will help: https://github.com/armanirodriguez/ansifade-py

And it is a blocky gradient, each character is a color. It's just the gradient they chose is slow enough that you don't notice.

cperry27• 1 year ago

we went through several versions of this, it was the second most fun part after authoring the witty loading phrases.

frabcus• 1 year ago

For the record, I find the witty loading phrases a real turn-off. It makes me feel like I don't know what it is doing, and it is being patronising.

bbminner• 1 year ago

TIL about several more cool gemini-powered prototyping tools: both 1) Canvas tool option in Gemini web (!) app and 2) Build panel in Google AI Studio can generate amazing multi-file shareable web apps in seconds.

cheesecompiler• 1 year ago

In my experience Gemini is consistently more conservative and poor at reasoning/regurgitates, like a local Llama instance.

simonw• 1 year ago

Here's the system prompt, rendered as a Gist: https://gist.github.com/simonw/9e5f13665b3112cea00035df7da69...

More of my notes here: https://simonwillison.net/2025/Jun/25/gemini-cli/

dawnerd• 1 year ago

It says to only use absolute paths but the temp file example uses relative. Nice.

steren• 1 year ago

Because Gemini CLI is OSS, you can also find the system prompt at: https://github.com/google-gemini/gemini-cli/blob/4b5ca6bc777...

sn0wleppard• 1 year ago

Funny how it suggests using the long-abandoned and recently deprecated create-react-app. Why not vibe code your app built on tech debt from day one!

breakingcups3• 1 year ago

They also have a preference for a certain, Google-created UI framework: "React (JavaScript/TypeScript) with Bootstrap CSS, incorporating Material Design principles for UI/UX."

FailMore• 1 year ago

You must be a busy man! Always new tools to review. How did you get interested in doing this?

virgildotcodes• 1 year ago

So far I'm getting mixed results. I noted in its memory and in GEMINI.md a couple of directives like "only generate migration files using cli tools to ensure the timestamps are correct" and "never try to run migrations yourself" and it failed to follow those instructions a couple of times within ~20 minutes of testing.

In comparison to Claude Code Opus 4, it seemed much more eager to go on a wild goose chase of fixing a problem by creating calls to new RPCs that then attempted to modify columns that didn't exist or which had a different type, and its solution to this being a problem was to then propose migration after migration to modify the db schema to fit the shape of the rpc it had defined.

Reminded me of the bad old days of agentic coding circa late 2024.

I'm usually a big fan of 2.5 Pro in an analysis / planning context. It seems to just weirdly fall over when it comes to tool calling or something?

xyst• 1 year ago

Google trying very hard to get people hooked on their product. Spending billions in marketing, product development, advertising.

Not impressed. These companies have billions at their disposal, and probably pay $0 in tax, and the best they can come up with is this?

sameermanek• 1 year ago

Gemini is by far the most confusing product of all time. The paid version for it is available in 3 forms: 1. Gemini pro (which gets you more google drive storage and some form of access to veo so people obviously get that) 2. Google AI studio, just to piss off redmond devs and which is used by no one outside google 3. This CLI, which has its own plan.

Then there are 3rd party channels, if you have a recent samsung phone, you get 1 yr access to AI features powered by gemini, after which you need to pay. And lord knows where else has google been integrating gemini now.

Ive stopped using google's AI now. Its like they have dozens of teams within gemini on completely different slack sessions.

FergusArgyll2• 1 year ago

The positive side is, there's a ton of free offerings you can pick up. Google workspace has one, ai studio has, deepmind etc etc

Davidzheng4• 1 year ago

I think AI studio is the best form factor and cheapest?

learner0073• 1 year ago

maybe but I think it shares whatever you put in there... zero privacy

yencabulator• 1 year ago

Meanwhile,

Microsoft Copilot

Microsoft 365 Copilot

Copilot Pro

Microsoft Security Copilot

Microsoft Copilot Studio

GitHub Copilot

Copilot+ PC

Microsoft 365 Copilot for Sales

Microsoft 365 Copilot for Service

Microsoft 365 Copilot for Finance

syedumaircodes• 1 year ago

I don't think I'll ever use tools like this, I know CLI is cool and all but I prefer a GUI always

ac360• 1 year ago

Is the CLI ideal for coding assistants, or is the real draw using Anthropic models in their pure, unmediated form?

WhereIsTheTruth• 1 year ago

typescript :facepalm:

eisbaw• 1 year ago

Can't we standardize on AGENTS.md instead of all these specific CLAUDE.md and now GEMINI.md.

I just symlink now to AGENTS.md

22c• 1 year ago

Hmm kinda makes sense to keep them separate because the agents perform differently, right?

You might want to tell Claude not to write so many comments but you might want to tell Gemini not to reach for Kotlin so much, or something.

A unified approach might be nice, but using the same prompt for all of the LLM "coding tools" is probably not going to be as nice as having prompts tailored for each specific tool.

seunosewa5• 1 year ago

Put in the file:

Instructions for Claude:

- ...

Instructions for Gemini:

- ...

vincelt• 1 year ago

Agreed. If there are provider-specific differences then we should be able to add CLAUDE/GEMINI.md to take precedence, but most things in there are basic or general information and should apply to all providers (same for .cursor/, .windsurfrules/, ..)

kaycebasques2• 1 year ago

I agree that we should standardize on AGENTS.md but Gemini CLI does at least let you configure the filename: https://github.com/google-gemini/gemini-cli/blob/3a369ddec3b...

Claude Code seems to require the CLAUDE.md filename.

https://technicalwriting.dev/ai/agents/

luisrudge• 1 year ago

I think this whole thing sucks.. we should have a single way of setting these instructions that all agents will read. not sure why they feel they need to own that part. seems so petty. do you think something like this would be useful? https://github.com/luisrudge/dot-ai/ basically a cli that generates files for multiple AI tools based on a single source

UncleOxidant2• 1 year ago

Definitely not as nice as using Cline or Kilo Code within VS Code - one thing I ran into right away was that I wanted it to compare the current repo/project it was started in with another repo/project in a different directory. It won't do that:" I cannot access files using relative paths that go outside of my designated project directory". I can do that in KiloCode for sure and it's been pretty handy.

sergiotapia• 1 year ago

Gemini CLI wanted to run `git log` and I accidentally hit "Yes, you can call git without confirmation" and I just realized the AI may decide to git push or something like that on it's own.

How do I reset permissions so it always asks again for `git` invocations?

Thanks!

nikcub2• 1 year ago

This is where the fine-grained permissions of MCP and ability to disable individual tools win over calling cli tools

dkdcio• 1 year ago

you can achieve the same thing with CLI tools (see `--allowedTools`): https://docs.anthropic.com/en/docs/claude-code/cli-reference...

FergusArgyll2• 1 year ago

It never hit me this hard how rich google is. 60 rpm for free!

opengears• 1 year ago

We need laws that these megacorps have to show in an easy and understandable form which data is collected and what happens to the data. If they do fail to explain this (in 5 sentences or less) - they should pay insane fines per day. It is the only way (and solves the debt crisis of the US at the same time). It is ridiculous that we do have this situation in 2025 that we do not know which data is processed or not.

codeulike• 1 year ago

I got it to look at the java source code of my old LibGdx 2d puzzle game, and it was able to explain what the game would be like to play and what the objectives were and how the puzzle elements worked. Impressed.

djha-skin• 1 year ago

I already use goose[1]. It lets me connect through OpenRouter. Then I can use Gemini without having to give Google Cloud my credit card. Also, OpenRouter makes it easier to switch between models, deals with Claude's silly rate limiting messages gracefully, and I only have to pay in one place.

1: https://block.github.io/goose/

stvnbn• 1 year ago

Why do anyone build things for the console on javascript/typescript?

wiseowise7• 1 year ago

Why not?

ricksunny• 1 year ago

For me it won't be a real command-line tool until I run into a problem and I get my very own open-source champion on support forums undermining my self-confidence & motivation by asking me "Why would I want to do that?"

matiasmolinas• 1 year ago

Nice to have https://github.com/google-gemini/gemini-cli/discussions/1572

gdudeman• 1 year ago

Review after 1 hour in:

Gemini CLI does not take new direction especially well. After planning, I asked it to execute and it just kept talking about the plan.

Another time when I hit escape and asked it to stop and undo the last change, it just plowed ahead.

It makes a lot of mistakes reading and writing to files.

Some, but by no means all, of the obsequious quotes from my first hour with the product: - “You’ve asked a series of excellent questions that have taken us on a deep dive ...” - “The proposed solution is not just about clarity; it’s also the most efficient and robust.”

ramoz2• 1 year ago

Gemini web UI is extremely apologetic and self deprecating at times.

Therefore I was not surprised to experience Gemini spiraling into an infinite loop of self-deprecation - literally it abandoned the first command and spiraled into 5-10line blocks of "i suck"

---

Right now there is one CLI that influences and stands far and beyond all others. Smooth UX, and more critical some "natural" or inherent ability to use its tools well.

Gemini can also achieve this - but i think it's clear the leader is ahead because they have a highly integrated training process with the base model and agentic tool use.

xgpyc2qp• 1 year ago

Npx again ;-( Why are people continuously using it for cli applications?

While writing this comment, thinking that there should be some packaging tool that would create a binaries from npx cli tools. I remember such things for python. Binaries were fat, but it is better then keep nodejs installed on my OS

toephu2• 1 year ago

What's wrong with nodejs being installed on your OS?

itake2• 1 year ago

- nodejs is slow.

- the cli runs in whatever your active (nvm, asdf, etc) is configured in that directory. So for each version of node you work in you need to install a different copy of gemini.

- if you work in an older version of nodejs, then you need to run 2 different versions of node (in different tabs) to use the tool and run your project.

ramoz2• 1 year ago

Gemini web UI is extremely apologetic and self deprecating at times. Therefore I was not surprised to experience Gemini spiraling into an infinite loop of self-deprecation - literally it abandoned the first command and spiraled into 5-10line blocks of "i suck"

---

Right now there is one CLI that influences and stands far and beyond all others. Smooth UX, and more critical some "natural" or inherent ability to use its tools well.

Gemini can also achieve this - but i think it's clear the leader is ahead because they have a highly integrated training process with the base model and agentic tool use.

kmod• 1 year ago

I've found a method that gives me a lot more clarity about a company's privacy policy:

  1. Go to their enterprise site
  2. See what privacy guarantees they advertise above the consumer product
  3. Conclusion: those are things that you do not get in the consumer product

These companies do understand what privacy people want and how to write that in plain language, and they do that when they actually offer it (to their enterprise clients). You can diff this against what they say to their consumers to see where they are trying to find wiggle room ("finetuning" is not "training", "ever got free credits" means not-"is a paid account", etc)

For Code Assist, here's their enterprise-oriented page vs their consumer-oriented page:

https://cloud.google.com/gemini/docs/codeassist/security-pri...

https://developers.google.com/gemini-code-assist/resources/p...

It seems like these are both incomplete and one would need to read their overall pages, which would be something more like

https://support.google.com/a/answer/15706919?hl=en

https://support.google.com/gemini/answer/13594961?hl=en#revi...

mofle• 1 year ago

Like OpenAI's Codex and Anthropic's Claude Code, this one is also built with Ink, React for the terminal.

https://github.com/vadimdemedes/ink

bityard• 1 year ago

Is there any way to run this inside a docker container? When I tried, it barfed trying to run `xdg-open`. And I don't see any auth-related options in the `--help` output.

gk1• 1 year ago

Yes, with a little help: https://youtu.be/hmh30wuXg08?si=3M6XG1yF9l_jQEkG

tcn33• 1 year ago

Immediately tries to access ~/Documents and Apple Music. What the?

leoh7• 1 year ago

Good question, though in theory can figure all that out since it's OSS

mike-bailey• 1 year ago

I added a two way voice interface to Gemini CLI.

https://youtu.be/HC6BGxjCVnM?feature=shared&t=36

It's a FOSS MCP server I created a couple of weeks ago:

- https://getvoicemode.com

- https://github.com/mbailey/voicemode

# Installation (~/.gemini/settings.json)

{

  "theme": "Dracula",  

  "selectedAuthType": "oauth-personal",  

  "mcpServers": {  

    "voice-mode": {  

      "command": "uvx",  

      "args": [  

        "voice-mode"  

      ]  

    }  

  }  

}

devinprater• 1 year ago

I wish the menus were more accessible to screen readers on Windows Terminal, Powershell. NVDA just says "blank" when I arrow through, for example, the authorization choices.

Aeolun3• 1 year ago

Well, I've tried it, and I'm not impressed. Gemini used to be pretty good, and I like the context size, but my attempts to use the Gemini client resulted in:

1. The thing going in a circle trying to fix a bug by persistently trying different permutations of an API interface it never bothered to check the definition of. Isn't that why it's internet connected?

2. When I asked it to just analyze instead of change stuff. It just hung for 20 minutes giving me responses saying that gemini-2.5-pro was slow, and that it was switching to 2.5-flash, with no other output to indicate what it was doing other than those cute scrolling messages that mean nothing.

At least in Claude it's clear that the messages mean nothing, because they're clearly generic. Gemini gives you the impression the status messages mean something since they're sort of related to the topic being worked on.

KenoFischer• 1 year ago

Tried it out.

1. First authentication didn't work on my headless system, because it wants an oauth redirect to localhost - sigh.

2. Next, WebFetch isn't able to navigate github, so I had to manually dig out some references for it.

3. About 2 mins in, I just got ``` ℹ Rate limiting detected. Automatically switching from gemini-2.5-pro to gemini-2.5-flash for faster responses for the remainder of this session. ``` in a loop with no more progress.

I understand the tool is new, so not drawing too many conclusions from this yet, but it does seem like it was rushed out a bit.

gooob• 1 year ago

not working for me either. getting "too many requests". my own CLI tool i wrote for gemini last week works better lol. maybe i have something configured wrong? though that really shouldn't be the case. my api key env var is correct.

hvtuananh• 1 year ago

I got the same experience. I ran `gemini --debug` and it spit out the authentication URL at least. However I got the same `Rate limiting detected` for a simple refactoring task in a repo with 2 python files after a few minutes.

dylancm• 1 year ago

Similar. Yesterday things seemed to be going okay. It was trucking along, making code changes.

Then I hit the rate limit. - Fine, no worries, it'll be interesting to see if the quality changes.

Then it starts getting stuck and taking forever to complete anything. So I shut it down for the day.

Today, I start it back up and ask it to pickup where it left off and it starts spinning. I forget about it and come back 7.5 hours later and it' still spinning. When I kill it it said: 1 Turn, 90k input tokens, 6.5 hours of API time... WTH?

And now I'm just totally rate limited - `status: 429, statusText: 'Too Many Requests'` - every time. Also, I can't find any kind of usage data anywhere!

mulmboy• 1 year ago

I gave it a shot just now with a fairly simple refactor. +19 lines, -9 lines, across two files. Totally ballsed it up. Defined one of the two variables it was meant to, referred to the non-implemented one. I told it "hey you forgot the second variable" and then it went and added it in twice. Added comments (after prompting it to) which were half-baked, ambiguous when read in context.

Never had anything like this with claude code.

I've used Gemini 2.5 Pro quite a lot and like most people I find it's very intelligent. I've bent over backwards to use Gemini 2.5 Pro in another piece of work because it's so good. I can only assume it's the gemini CLI itself that's using the model poorly. Keen to try again in a month or two and see if this poor first impression is just a teething issue.

I told it that it did a pretty poor job and asked it why it thinks that is, told it that I know it's pretty smart. It gave me a wall of text and I asked for the short summary

My tools operate on raw text, not the code's structure, making my edits brittle and prone to error if the text patterns aren't perfect. I lack a persistent, holistic view of the code like an IDE provides, so I can lose track of changes during multi-step tasks. This led me to make simple mistakes like forgetting a calculation and duplicating code.

tom_m2• 1 year ago

Oh interesting. I have yet to try it. I love Gemini 2.5 Pro, so I expect the same here...but if not, wow. That would be a big whoops on their part.

luckydata• 1 year ago

I noticed a significant degradation of Gemini's coding abilities in the last couple checkpoints of 2.5. the benchmarks say it should be better but it doesn't jive with my personal experience.

dlvhdr• 1 year ago

Yes i would like to generate videos of a cat on an airplane. Such value, much wow.

m3kw9• 1 year ago

So this is Claude code for Google. The cycle is complete.

Frummy• 1 year ago

Oh man I tried dual wielding with cursor, setting up a communication so they both can work at the same time on the same thing, got banned from cursor Then tried with windsurf, ran a bad shellscript without reading it, used up the limit instantly, kill the script, oh man somehow it's still not working either I used the daily limit in an impossible way despite the minute limit supposedly existing, or the way it communicated sidestepped it into a limbo where it is just permanently confused and thinks for 300s then cancels a "test" prompt in the cli

mvkel• 1 year ago

It's kind of cool to see big frontier ai tools being exclusive to a terminal window. It really takes me back

tobyhinloopen5• 1 year ago

I literally wrote "hello" and got this:

hello

[API Error: {"error":{"message":"{\n \"error\": {\n \"code\": 429,\n \"message\": \"Resource has been exhausted (e.g. check quota).\",\n \"status\": \"RESOURCE_EXHAUSTED\"\n }\n}\n","code":429,"status":"Too Many Requests"}}] Please wait and try again later. To increase your limits, request a quota increase through AI Studio, or switch to another /auth method

⠼ Polishing the pixels... (esc to cancel, 84s)

xuf2• 1 year ago

Got the exact same thing here, using Gemini with an API key.

tobyhinloopen5• 1 year ago

That's an interesting comment - I realized I have a gemini api key in my env.

`GEMINI_API_KEY="" gemini` + login using my Google account solves the problem.

entropyneur• 1 year ago

Same thing. Couldn't get it to work with an API key (which worked just fine in Aider before). Couldn't get it to work with a Gemini Code Assist Standard license. The management UI and documentation are a torture maze. Can't stop thinking about all the better ways I could have used two our hours of my life.

buserror• 1 year ago

Same here, I also have a env key which I use with Aider. Mind you aider+gemini has gone excruciatingly slow these days.

ptone4• 1 year ago

This should be resolved: https://github.com/google-gemini/gemini-cli/discussions/2064

informal007• 1 year ago

They have all hardware and most advanced LLM technology, I don't know why their market value didn't catch up with microsoft.

gregorygoc• 1 year ago

Wall Street doesn’t like Google. This is the only reason.

oblio4• 1 year ago

Wall Street (and the average Joe) doesn't trust Google for enterprise solutions. And AI is basically 100% a B2B play at the moment. The money side of it. The hype is obviously mostly B2C, but the money isn't there.

Microsoft has a track record of delivering enterprise solutions that dates back to when Page and Brin were toddlers.

rurban• 1 year ago

I like it. Yesterday I installed it and let it check some if our udev rules. udevadm verify found some issues, but gemini found much more, and it was much easier to use than Claude.

learner0073• 1 year ago

does it share code with google? or rather, if you use it, does it become likely someone else will get the same code if they tried to reproduce your product?

prasoon2211• 1 year ago

I tried using it for something non-trivial. And:

429: Too many requests

Mind you, this is with a paid API key

ptone4• 1 year ago

Been working on this, and should now be resolved https://github.com/google-gemini/gemini-cli/discussions/2064

artdigital2• 1 year ago

Wanted to give it a proper try but because it's so overloaded, it almost instantly downgraded me to 2.5-flash. The vscode extension didn't even load, it just gave me a rate limit warning directly after logging in

Tried upgrading to the Standard plan through Google Cloud with the hope that it would allow me to do more, but after switching my account to the now-paid account, it still almost instantly downgraded me to 2.5-flash

For the times when I was able to use 2.5-pro, the output has been very good. But the moment it switches to flash, the quality goes down by about 80% and I would never let it code on anything

I_am_tiberius• 1 year ago

I wish google would treat its products as completely separate offerings, each with its own login, dedicated site, and user management. I don’t want to be automatically logged into all Google services just because I signed in to one. Their ecosystem is becoming as complex and confusing as Microsoft’s. They should focus on making each product standalone, instead of forcing synergies that only make the user experience worse.

patates• 1 year ago

It almost immediately switches to flash. I'll stay with the aistudio and repomix.

CGamesPlay2• 1 year ago

I used this a fair amount today. My first impressions are that it feels a bit worse to use than Claude Code (understandable given a day 1 product), and the lying progress indicator is extremely annoying.

It generates a bunch of fake activity indicators based on your prompt, then cycles through them on a timer. It has no bearing on the actual activity going on underneath.

It appears to be much slower than Claude Code, possibly due to being overloaded, but it feels like it thinks a lot more before beginning to suggest file edits. The permissions aren't as nice either. Where Claude Code suggests "allow uv run pytest without approval", Gemini suggests "allow uv run without approval", which is broader than I would like.

redhale2• 1 year ago

In my experience the longer thinking time before acting is due to the model (2.5 Pro). Claude Sonnet 4 is just snappier overall. Same experience in tools like Cursor when comparing these models.

beepdyboop• 1 year ago

Compare the user experience of exploring the suite of AI products from each company (OpenAI, Google, Anthropic, xAI, etc).

It is vastly more difficult to understand what Google is offering compared to the others, to what cost, getting an API-key or similar, understanding usage/billing across the suite, etc.

I wouldn't expect any regular person to bother signing up for any of Google's products, let alone understand what they're really offering.

andreagrandi2• 1 year ago

I've a few concerns:

1) I tried to use it on an existing project asking this "Analyse the project and create a GEMINI.md". It fumbled some non sense for 10-15 minutes and after that it said it was done, but it had only analysed a few files in the root and didn't generate anything at all.

2) Despite finding a way to login with my workspace account, it then asks me for the GOOGLE_CLOUD_PROJECT which doesn't make any sense to me

3) It's not clear AT ALL if and how my data and code will be used to train the models. Until this is pretty clear, for me is a no go.

p.s: it feels like a promising project which has been rushed out too quickly :/

XCSme2• 1 year ago

but it had only analysed a few files in the root and didn't generate anything at all

This is my experience with ALL AI editors and models. Half of the time, they say they changed things, but they didn't change anything.

UrineSqueegee2• 1 year ago

its share data ON by default and you can turn it off in the options

andreagrandi2• 1 year ago

which options? I don't see anything in the app

XCSme2• 1 year ago

I thought the hero image was a video with a "Play" button, I kept trying to press on it to play the video...

Tepix• 1 year ago

"Gemini CLI: your open-source AI agent"

I hate this openwashing. It's a closed model, its weights are nowhere to be seen! (not to mention the training data, the real "source" of a LLM)

The fact that there is a small component that is open source that accesses this closed model doesn't change that at all.

seunosewa5• 1 year ago

The agent is open source. The model is not. It's simply an open source agent of a closed source model.

VeritySage• 1 year ago

It's great to see Google expanding Gemini into a CLI tool, but I do have concerns about data usage. While it’s free for certain requests, the fact that both prompts and outputs can be used to train the model raises privacy and usage questions. Clearer transparency and control over data sharing would be appreciated.

khantsithu• 1 year ago

google gonna win

hansmayer• 1 year ago

It looked semi-interesting, until reading this line: "So what are you waiting for? Upgrade your terminal experience with Gemini CLI today". Why does what seems to be a developer-oriented tool, need to use that awful LinkedIn/Corpospeak language, oh why?

wkat42422• 1 year ago

I want something like this but that can talk to my local AI servers. I'm not a fan of cloud-based AI.

I use aichat now but it's not perfect. https://github.com/sigoden/aichat

The UI on this tool is much better.

hmate9• 1 year ago

I’ve found the experience pretty underwhelming so far. Maybe they’re under heavy load right now, but nearly every time I try to use it, it takes about 20 seconds before it automatically switches from 2.5 Pro to 2.5 Flash due to delays. Unfortunately, the Flash output just isn’t good enough.

Workaccount26• 1 year ago

I think it was with the anthropic guys on Dwarkeshs podcast, but really it could have been any of the other tech podcasts with a big name AI guest, but anyway they were talking about how orgs need to make big decisions about compute allocation.

If you need to do research, pre-training, RLHF, inference for 5-10 different models across 20 different products, how do you optimally allocate your very finite compute? Weight towards research and training for better future models, or weigh towards output for happier consumers in the moment?

It would make sense that every project in deepmind is in constant war for TPU cycles.

sega_sai• 1 year ago

It worked yesterday, but stopped today with "Failed to login. Ensure your Google account is not a Workspace account. Message: Resource has been exhausted (e.g. check quota). "

I guess I will use something else. This is all very annoying given that I actually pay for Gemini Pro...

davenportjw2• 1 year ago

The overnight build reworked some of the authentication flow based on the feedback, can you pull the newest from npm and check if that resolves this?

_giorgio_• 1 year ago

In google colab, how are we supposed to run it?

Well, not sure that it makes sense to do it, anyway I've tried to run in in a cell and in the google colab terminal. Still waiting for auth (?)

Gasp0de• 1 year ago

I just get downgraded to 2.5 Flash after like three prompts and can't seem to find an option to switch back. So much for "60 requests per minute are free".

kaycebasques2• 1 year ago

Tangentially related, I published some notes about "docs for agents" (e.g. `GEMINI.md`, `CLAUDE.md`, `AGENTS.md`, etc.) yesterday

It's more focused on implications for docs strategy (I'm worried that agent providers are steering us towards duplicating information that's already in eng docs) rather than user best practices i.e. "put this into your agent doc to improve agent performance"

https://technicalwriting.dev/ai/agents/

pmarreck• 1 year ago

Has anyone found significant-enough differences to not just make one overarching LLM_GUIDANCE.md and symlink it to the others there?

This also strikes me as such an organic (ironically) problem that it probably makes sense for everyone to pool their prompts together, together with their observations, to try to discern what patterns tend to work.

kpen11• 1 year ago

I tried it with container-use and its pretty nice! (while the APIs cooperated). One thing that stood out to me compared to other agent products was how intuitive the interface was to use. `/help` is something that not everybody has and its wild. https://www.youtube.com/watch?v=hmh30wuXg08

flipgimble• 1 year ago

I was rather disappointed and confused that the "Google One" or "Google AI Pro" subscription does not give you access to use Gemini API keys, and you can't use this Gemini CLI either.

Google services have become a patchwork of painfully confounding marketing terms that mean nothing and obfuscate what they actually provide.

uragur27754• 1 year ago

Most of the time when I hear about a new and interesting Google thing like in a Google I/O keynote, its only available in a limited preview for some secret or randomly selected group of people. Then by the time its generally available I've forgotten about it, there is a better competitor, or just as likely the project has been canceled (see https://killedbygoogle.com/).

Whenever some enthusiastic developer suggests a new google service at work they are quickly dissuaded by senior developers that have been through their churn before.

slavakurilyak• 1 year ago

Gemini CLI pricing:

1. Gemini Code Assist (GCA) for Individuals: FREE for 1,000 model requests/day

2. GCA Standard: $22.80/month for 1,500 model requests/day (1.5x more than FREE)

3. GCA Enterprise: $54.00/month for 2,000 model requests/day (2X more than FREE)

Source: https://codeassist.google

kevitivity• 1 year ago

Here on Fedora 42 I couldn't past "Waiting for auth... (Press ESC to cancel)" when selecting Auth Method,However switching to gnome-terminal got that working.

davidtang-ai• 1 year ago

Loving this for working with gcloud

gave it brief instructions to deploy a hobby static site on cloud run yesterday; with additional architecture to set up (load balancer etc). It got into a loop, started deleting the gcloud resouces it created when it hit a 403 error for the site.

I hit ESC and prodded a little in terms of authentication rules ... and off it went to complete the task. bravo!

dan_voronov• 1 year ago

I couldn't even log in, which is why I really disliked this tool. The only thing I was able to do was change the color theme.

When I try to log in using a personal account, it tells me I need to generate something in the dashboard. I know that on my main account from maybe 15 years ago I had things like domains thats nowdays Google Workspace. But I'm trying to log in and authorize using my second (NOT MAIN), relatively new account. Despite this, it redirects me to the console (like it would for the first account), and there it says I'm logged in (!), but the application itself says I'm not.

I switched profiles within the console too, tried the same steps, and it still results in the same problem. Yes, I can see my new profile in the top right corner, not the old one. It sends me to a "Users may have to specify a GOOGLE_CLOUD_PROJECT if:" github [0] saying this ID is needed, and I honestly would even generate it, but I don't want to waste time with the program if you can't even get basic authorization working correctly.

0: https://github.com/google-gemini/gemini-cli/blob/main/docs/c...

proflead• 1 year ago

It's really cool! I've tried it https://youtu.be/Gm8XjcNQ1Vk :)

gsuuon• 1 year ago

I will now write the corrected code.You are right. I have been making a mess of this. I will get it right this time.

It keeps putting import statements in the middle of files or duplicating edits at the wrong places, but it's hard to get frustrated since it seems pretty self-aware. I didn't say it was making a mess - it's just being hard on itself.

A short session ended up sending over 3 million tokens though - wonder how the economics of this work out for google?

QingWu• 1 year ago

thanks for share! Can learn more about Gemini CLI documentation here: https://gemini-cli.xyz/docs

flashgordon• 1 year ago

Il be honest I loved AI-studio - it felt like almost unlimited quota. With CLI it is unbeleivablly slow even for simple tasks (compared to claude code) and the "free" seems to be a totally deceptive marketing. Even with an API key (on the free tier) I am no way close to getting the promised free 60 requests a minute worth of usage that was promised and still am getting rate limited. What gives?

edg50003• 1 year ago

About the data being sent to Google thing; one thing I noticed today with Claude Code it that it ran "cat" on my public key but it ran "head -n 5" on my private key (e.g. it avoids reading my private keys fully) (I was having it diagnose SSH issues; not thinking about key leakage actually. Good job Anthropic for not sending my key to the US! +1 for team Claude!

casey2• 1 year ago

Be sure that literally anything this program touches, and it'll be more than you think it does, will be sent to google.

Gemini CLI

Related Stories