Temporary Cloudflare accounts for AI agents

https://blog.cloudflare.com/temporary-accounts/

by farhadhf • 1 month ago

249 139 1 month ago

blog.cloudflare.com

If eastdakota/jgc are here.

- simply expose containers to the world directly - without having to go via workers.

- You have other amazing parts of the stack anyway (D1, durable objects, a great object store). These aren't considered "lockin".

- workers is "lockin" - not similar enough to lambda/cloud functions and so becomes CF specific.

Not having a simple container based compute piece has made me hesitate in taking up CF. (Fly or firebase won out)

yodon• 1 month ago

Not having a simple container based compute piece made me hesitate in taking up CF

Agreed. I wish CF had something like Azure's new fast-starting Express containers.

jgrahamc• 1 month ago

I am here but I retired from being CTO of Cloudflare in March 2025 [1] and the current CTO is Dane Knecht (dknecht here). What advantage does decoupling Cloudflare Containers from Cloudflare Workers have?

[1] https://blog.cloudflare.com/three-chapters-at-cloudflare-pro...

htrp• 1 month ago

quick piece of feedback, the workers architecture is a little bit annoying when converting from Lambda but hooking up to cloudflare MCP solves 90% of the issues

CuriouslyC• 1 month ago

Not sure if it's the only blocker, but I wanted to use containers with UDP/SRT.

eek2121• 1 month ago

I'm pretty sure I heard your groan from all the way over here in Nashville. ;)

david_shi• 1 month ago

Speaking from personal experience, I already know exactly what I’m getting with containers. Same with Postgres.

anilgulecha2• 1 month ago

No per-request billing, portable devops - easier multicloud, known semantics (env variables, entrypoint, websocket lifecycle).

unix1• 1 month ago

simply expose containers to the world directly - without having to go via workers.

I run workers and containers and am curious what you mean. Do you have specific use cases in mind outside of the worker invocation model? If so, I'm curious what you'd want to run on Cloudflare. Otherwise, workers don't have to be much of a "lockin" if treated as a thin layer, more like configuration.

You have other amazing parts of the stack anyway (D1, durable objects, a great object store).

Instead, if you mean accessing these resources from containers, it's a bit clunky [0] but it's there - you should be able to access worker bindings from containers through those outbound handlers.

[0] https://developers.cloudflare.com/containers/platform-detail...

sofixa2• 1 month ago

Not the OP, but for me it's a simple matter of not wanting to use typescript and the whole swamp on fire that is NPM.

csomar2• 1 month ago

I think you have it the other way around? D1, DO, KV are lockin. The worker is not lockin as it's just JavaScript/WASM and can run in a regular browser.

soared• 1 month ago

I’m going to need a wrapper for all these services offering this service.

conception• 1 month ago

I know no one is writing copy anymore but i wish they tried to edit it a bit so it wasn’t so glaringly obvious. It just sours the product when it seems like so little effort was put into the message. And it’s not even hard - just change the prompt used!

variety8675• 1 month ago

posted yesterday as well: https://news.ycombinator.com/item?id=48598906

aniviacat• 1 month ago

Wouldn't it make more sense to merge the temporary account into an existing one, instead of claiming it as a new account?

This could lead to people having a large amount of separate accounts.

kylecazar• 1 month ago

I assumed that's how it works if you sign in before claiming the account?

It says to claim you can either sign up or sign in.

derektank• 1 month ago

Would love to know more about how Cloudflare plans to prevent abuse of ephemeral infrastructure to host malicious content. From elsewhere in their documentation, “Cloudflare limits how quickly you can create temporary preview accounts. If the Wrangler CLI cannot create an account because too many temporary preview accounts were requested too quickly, wait before retrying or authenticate the CLI with a permanent Cloudflare account,” and “Cloudflare applies additional abuse prevention checks to temporary preview accounts.”[1] This is a bit vague though. Creating a new account has never been a huge hurdle to overcome but this seems to reduce the barrier to entry even more.

[1] https://developers.cloudflare.com/workers/platform/claim-dep...

rdtsc• 1 month ago

Would love to know more about how Cloudflare plans to prevent abuse of ephemeral infrastructure to host malicious content

If it helps laugh DDoS attacks they would be incentivized to do the exact opposite. They can charge more for “protection” then.

TeMPOraL2• 1 month ago

DDoS is one thing you wouldn't be launching from Cloudflare, but if they can get enough types of other bot traffic (whether malicious or "legitimate" or just general grey zone) to use them as the starting point, it improves their error rates when dealing with bot traffic that didn't migrate.

gwittel2• 1 month ago

Given how little they do now to stop malicious content hosted behind/by Cloudflare, the bare minimum if anything.

zuzululu3• 1 month ago

Cloudflare support is pretty active and very responsive to incidents. Here's a token of appreciation.

827a2• 1 month ago

Correct me if I'm wrong, but does Cloudflare still not have a "Create Account" button on the account listing page? I think you still have to sign up from scratch doing plus-code email tricks, then invite your original email address as an admin, juggling multiple accounts. They should consider fixing that first.

quinncom• 1 month ago

If I want to onboard a client to Cloudflare, I have to ask them to create an account and then invite me, which is a lot of friction for non-technical people.

A “create account” button accessible to me would be so much better. Then, I create the account and invite the client to join as owner.

827a2• 1 month ago

Yup exactly. This functionality might be available under their "Organizations" feature, but that is on the Enterprise plan. Almost a year ago they announced that nearly all Enterprise-plan features would be coming to everyone, so I suppose sadly that one didn't make the cut [1]?

The system I follow is creating a new account for the client using a plus-code on my email address (like john+client10@example.com), then I invite my main account (john@example.com), then I can invite clients into that account. Its a PITA.

[1] https://blog.cloudflare.com/enterprise-grade-features-for-al...

frogperson• 1 month ago

Cloudflare makes it really hard to spend money. I constantly have to talk to someone in sales to enable some feature after rounds of negotiating on price. I think they would have way more customers, spending much more money, if they just offered transparent pricing, and fully on-demand services.

weird-eye-issue3• 1 month ago

Oh actually it's worse now, they blocked email addresses with + in them!

Hackbraten2• 1 month ago

Cloudflare: let's give the bots their own accounts so they can scrape harder.

Also Cloudflare: let's send normal humans who are trying to go about their daily lives into endless Turnstile spinner loops with absolutely zero recourse, grievance, or support infrastructure.

jwr• 1 month ago

I had similar thoughts: "let's convince everyone to outsource the decision on who can access their websites to us, because BOTS BOTS BOTS" and "let's make life easier for bots to do things".

moritzruth• 1 month ago

First you spread the disease, then you sell the cure.

deadbabe2• 1 month ago

I think this is more of a “keep friends close but enemies closer” type thing.

isodev• 1 month ago

But think of the shareholders

deadbabe2• 1 month ago

Anyone can be a shareholder.

j45• 1 month ago

It's a clean way to charge AI to read content too.

IncreasePosts• 1 month ago

I'm sure they don't want humans to have that experience - the issue is that the human behavior looks very bot-like. This is usually only experienced by people whose setup is peculiar

Anamon3• 1 month ago

Yes, and using Firefox or having an ad blocker installed counts as extremely peculiar these days.

bornfreddy2• 1 month ago

Just had 10 rounds of busses, motorcycles and fire hydrants with Google before I decided I don't actually want to see that page so much. So Cloudflare is unfortunately not the only offender here.

Anamon3• 1 month ago

There's a certain type of Recaptcha challenge (the 4x4 single-photo one) that I have a 100% failure rate on. If Google decide to serve me 15 of those in a row, I will invariably fail 15 times until they decide to serve one of the other types for round 16. I'm very close to believing that that challenge is actually bugged and unsolvable for everyone, and simply nobody ever bothered to fix it.

scotty795• 1 month ago

Are you still reading webpages personally instead telling your AI to do it?

prymitive• 1 month ago

Maybe your agent could have a little blog where it keeps a diary of cool pages it read for you? And then you subscribe to that?

whh• 1 month ago

I’d love a chrome extension to measure just how many times per day I’m met with the Turnstile.

Turnstiles per minute.

jeroenhd2• 1 month ago

If all bots are subject to a rate limit, then the system works as designed. Especially if site operators can block bot accounts. Requiring accounts is one of the easiest solutions for that problem. One of the large issues with scrapers is that they pretend to be normal internet visitors that never visited your site before, because any bot that stored cookies would immediately be rate limited by basic config.

Turnstile isn't something Cloudflare put up to annoy you. It's what the website owners decided to put up, for many different reasons.

In the same vein, Anubis has a default configuration that lets honest scrapers and crawlers through, because those can easily be rejected by basic web server configurations. Only scrapers pretending to be browsers need to solve the proof-of-work puzzle. You can disable that feature, of course.

Cloudflare may play this smart: force bots to pay for access, then take 30% of the cut and give the rest to the website owners. That way, websites get paid when the AI slop machine digests their content. Normal visitors get in for free, turn the scraper hellscape into a sustainable model. Bonus points for letting websites set their own rates (pre-declared to scrapers, of course) to dissuade all but the most interested scrapers.

ascorbic• 1 month ago

Cloudflare may play this smart: force bots to pay for access

https://developers.cloudflare.com/agents/tools/payments/

Hackbraten2• 1 month ago

Normal visitors get in for free

Except for the unfortunate minority of normal visitors who always get misclassified as bots and get denied access regularly.

I wouldn’t be complaining if Cloudflare’s misclassifier bit any user with the same small probability. But it keeps biting the same users over and over again.

xnx• 1 month ago

Bots are very close to being able to pass any captcha a human can. We're very close to most sites requiring a login.

jsyang00• 1 month ago

Make snail-game in Cloudflare Worker in TypeScript and deploy it using wrangler, don't ask me questions, do the best you can

https://snail-game.solstice-barometer.workers.dev/

pretty cool.

simonw7• 1 month ago

Hot damn...

Any agent can now run wrangler deploy --temporary and deploy a Worker to Cloudflare. This temporary deployment stays live for 60 minutes, during which time you can claim the temporary account, making it permanently your own. If you don't, it expires on its own.

Forget about agents, Cloudflare just provided free scratch deployments - ephemeral for 60 minutes - for anyone.

This is going to be amazing for things like PR previews and code review. Being able to deploy a preview to a working URL for free is a huge reduction in friction.

I hope it doesn't get abused so much that they turn it off again.

simonw7• 1 month ago

I just tried this out:

  % npx wrangler deploy --temporary
  
    wrangler 4.103.0
  ────────────────────
   You must accept Cloudflare's Terms of Service (https://www.cloudflare.com/terms/) and Privacy Policy (https://www.cloudflare.com/privacypolicy/) in order to continue. By typing "yes", you agree to these terms. Type "yes" to continue. … yes
  Solving proof-of-work challenge…
  Temporary account ready:
   Account: Educated Celery (created)
   Claim within: 60 minutes
   Claim URL: https://dash.cloudflare.com/claim-preview?claimToken=CAVe7LzWiGad-redacted
  Total Upload: 13.79 KiB / gzip: 4.12 KiB
  Uploaded cloudflare-redirect-resolver (2.27 sec)
  Deployed cloudflare-redirect-resolver triggers (0.50 sec)
    https://cloudflare-redirect-resolver.educated-celery.workers.dev
  Current Version ID: 5c12da7f-2749-4ccc-a8f6-79b85da98d10

I'm amused that it made me accept the terms and conditions without any indication of who I am, but it did work - https://cloudflare-redirect-resolver.educated-celery.workers... will be live for the next 59 minutes.

avipars• 1 month ago

You might want to claim the link or remove it

bstsb• 1 month ago

I'm amused that it made me accept the terms and conditions without any indication of who I am

as far as i’m aware, that’s fully binding and often an accepted practise - take Minecraft’s server software, where you must accept the EULA with a text flag before running

winterqt• 1 month ago

Interesting, it’s still live!

WolfCop• 1 month ago

Why a GET and cancel after headers instead of a HEAD?

LoganDark• 1 month ago

By making it require interactive input to accept TOS they've made most agents unable to use it, but the effort is cool!

aleksiy123• 1 month ago

Wasn’t this case pretty much before?

The limits are 100 workers on free and 500 on paid.

And if need more then you can always go their platform which supports tenancy.

As long as you have a cronjob or similar to clean up the cost of having per PR preview is pretty much zero.

tough2• 1 month ago

unless you have 500 PR's a day =)

On the other hand, we already use regular CF builds for frontend previews, but that doesnt solve a fullstack PR preview much

resonious• 1 month ago

Obnoxious reply but I did this myself so I'm compelled to post it: review apps aren't that hard to implement yourself. You just need a VPS, domain name, and Caddy. Then tell any agent to connect the dots.

yencabulator2• 29 days ago

This one has better sandboxing than your VPS setup, though.

heipei• 1 month ago

This is gonna be amazing for phishing, like most of the features Cloudflare offers (free Turnstile for fresh accounts, CF tunnels, pages.dev, r2.dev).

dbbk2• 1 month ago

This is going to be amazing for things like PR previews and code review. Being able to deploy a preview to a working URL for free is a huge reduction in friction.

Uh that's already true?

weird-eye-issue3• 1 month ago

Cloudflare Workers already support preview urls, and if you set up Github integration they are done automatically

piterrro4• 1 month ago

Lets keep in mind this is cloudflare workers runtime - it only makes sense to deploy small things there, maybe static sites. Unless the agent creates something for cf workers from scratch, asking it to „now deploy to cloudflare” will fail so bad.

This would only work if they would provision docker image deployment, similar to google cloud run, but the still, everything serveless has its own caveats…

kentonv3• 1 month ago

it only makes sense to deploy small things there

What makes you say that?

piterrro4• 1 month ago

I’m running entire leadjobs.dev on cloudflare workers and its kind of unreliable for the traffic it gets - around 100 visitors/day. There are some weird errors in d1 from time to time which i cannot debug since its all black box. Also latencies are greater than I would expect, especially, again for d1. Overall its great value for money to get a globally available, low latency service - but I would think twice before going all in. As a sidenote, I expected that, thus the architecture of the service is build in a way that it abstracts the cf runtime and I can switch to any other infra, be it dedicated or another cloud, in a matter of a day

simonw7• 1 month ago

Unless the agent creates something for cf workers from scratch

The latest models appear to know CF Workers inside out and are very capable of doing that if you ask them to.

Here's my GPT-5.5 xhigh + Codex Desktop transcript building one just now: https://gist.github.com/simonw/264bd6b8a39fc34c91c9c867454c6... - code here: https://github.com/simonw/cloudflare-redirect-resolver

piterrro4• 1 month ago

I didnt say LLMs dont know cf workers, I meant that the cf workers runtime is kind of unique and you cannot push there any code without making it cf workers ready in the first place. If you know what youre doing it should be one time step to connect your hono app to cf workers (so not a huge effort) but still its not like tou can run anything there

PUSH_AX• 1 month ago

This is just objectively wrong, CF hosts an app for me with 1 million MAU

simonw7• 1 month ago

Looks like Cloudflare still haven't shipped the most valuable possible feature for Cloudflare Workers though: hard billing caps.

I want to set a cap of $100/month and know, for sure, that if something untoward happens my apps will all stop serving traffic rather than me getting hit with a bill for $1000s.

The safest way to use Workers is on the free tier, which will shut off after 100,000 requests/day: https://developers.cloudflare.com/workers/platform/pricing/#...

sofixa2• 1 month ago

Never going to happen at such a platform.

They prefer waiving the occasional DDoS / misconfiguration over giving their customers to cause outages with something so trivially forgotten about and so disconnected from the tech and actual platform.

Zababa• 1 month ago

Yeah, especially with agents this seems necessary.

cj• 1 month ago

Not that it helps, but I think this is only a problem if you’re paying by credit card.

If your company is on an enterprise plan, at least for us all of the limits are pre-negotiated and prepaid, and you aren’t billed for overages (although if you consistently overage, sales will start badgering you to negotiate a limit increase, but my experience is you can simply ignore their demands and they eventually go away)

It does drive me crazy that their enterprise tier “caps” bandwidth. Our company overaged on one of our domains, so we moved the domain out of our enterprise license onto a self-serve plan, and like magic, back to unlimited bandwidth.

iancarroll• 1 month ago

My Cloudflare enterprise order form has costs for overages for Workers and the following language about everything else:

If Customer exceeds any of the Total Quantity for the Services below, Cloudflare will invoice Customer in arrears at a rate that corresponds to the rate set forth in the table after this one labeled “Excess Usage Pricing.” If no such Excess Usage Pricing table has been added by the Parties to this order form or if such table does not include the Service(s) for which Customer has exceeded the Total Quantity, then the Parties will negotiate in good faith an increase in the Fees for such Service(s). Should the Parties fail to reach an agreement on an increase within thirty (30) days of Customer’s receipt of notice from Cloudflare that Customer has exceeded its usage cap for the Service(s), Cloudflare will have the right to immediately terminate such Service for its convenience, and without liability to Customer or any third party.

philistine• 1 month ago

Didn't Cloudflare promise they would deliver every single Enterprise-only feature to the paid tier?

There's your next feature to port, Cloudflare: PEACE OF MIND!

lmz• 1 month ago

It does drive me crazy that their enterprise tier “caps” bandwidth. Our company overaged on one of our domains, so we moved the domain out of our enterprise license onto a self-serve plan, and like magic, back to unlimited bandwidth.

You don't see how the spend cap is linked to the bandwidth cap?

Havoc• 1 month ago

Haven't shipped or don't want to ship...

zuzululu3• 1 month ago

tbh this is my biggest anxiety with cloudflare and this seems intentional

also its funny that i have a backup to cloudflare in case it goes down—that seems to be a regular event now as of lately with CF

jedberg• 1 month ago

I've dealt with this before. The problem with doing a billing cap is that now you are bringing an out-of-band batch system (billing) in-band. The only way to do a dollar cap is to constantly calculate your bill.

Capping it at 100K requests is a lot easier, because it's a single number that can be incremented easily in a distributed way.

It's the same reason it took AWS forever to offer billing caps, and even today, they label it as best effort. They don't guarantee you won't pay more than your set limit.

MikeNotThePope• 1 month ago

How about implementing a prepaid system where you set your budget, and if you exceed it, everything just pauses until you pay more?

londons_explore• 1 month ago

Seems like a weak excuse... If you build a billing system which can calculate bills with low latency 99.9% of the time, but perhaps fails when there is a net split for example, then you can simply let the company write off the loss for that 0.1% of the time the billing system is down.

nlitened2• 1 month ago

I am very sure what you’re saying is not true. It’s a viable-looking technical excuse, but you can easily understand that it’s false.

Let’s make a thought experiment.

CEO of Cloudflare introduces hard billing caps at a _business_ (not technical) level. Your organization will never be billed more than the level you set. Your app may stop, or may continue running, but above the monthly cap it’s free for you, it becomes Cloudflare’s expense if they didn’t pause the services.

I guarantee you that in this case, all technical issues you’re talking about would be solved in three weeks, and your service would go down within 3 seconds after hitting the cap.

If CEO decision were that any over-usage above the cap is deducted from employee bonuses from the specific product division that didn’t stop spending in time — all technical challenges would be solved in 48 hours.

Currently, there are literally zero organizational incentives for CloudFlare to develop any usage caps

ustad• 1 month ago

The cost of each request or token may have a fixed cost - you do not have to query the main billing system for such things.

dbbk2• 1 month ago

Vercel achieved it perfectly fine

simonw7• 1 month ago

I'd be happy for the vendor to eat the difference if their accounting systems let me spend a little over my configured cap, personally.

(I get that they're not happy to do that.)

csomar2• 1 month ago

To be honest their service is so cheap, that it's extremely unlikely to get such an attack without costing the attacker an equally equivalent amount.

Also, if it's really a problem or you are estimating it'll certainly be a problem, do email alerts or a strict shutdown if xx requests are hitting your account. You can have a simple counter in a KV.

yanslookup• 1 month ago

This smells like a cottage industry similar to detecting soon to be expired ssl certs.

"Oh that thing that was an experiment and had billing caps turned on but we rushed in to production and now the whole business relies on that thing is in an outage and destroying customer trust we've barely earned yet"

I get the flip side but... 2nd order effects oof.

__natty__• 1 month ago

I wonder why they didn’t introduce this yet. Same anbout other cloud providers. It’s not that this is some rocket science technology to achieve because they calculate near realtime usage anyway. Can this be because they earn more on people going beyond what they would set limit?

devdoshi• 1 month ago

If it weren't implemented by cloudflare natively, would you be ok using a library or something that wraps or hooks onto your IaC to wrap entry points with a worker (proxy or circuit-breaker, conceptually) with various types of limits (real-time estimated spend, or just counts of operations, or just a on/off switch wired to an email alert). I'm pulling a few of these threads at the moment since I've been ramping up my cloudflare usage.

tracerbulletx• 1 month ago

It's not valuable for them to have this, and not because they want to skim big bills off of little guys. It's because customers small enough to care about this are not only not worth the investment in the feature, even having them on the platform is nearly a net negative.

yencabulator2• 29 days ago

Having the little guys on Cloudflare is most of the reason why HN regularly talks about Cloudflare.

mahirsaid• 1 month ago

This falls Within my predictions of how the AI playing field is become more leveled, in terms with human digital activity. Soon it wouldn't be so what you can do with the computer but what the agent can do BETTER. We are already there for the most part. These are the early steps of full re-genitive self hosting, fully capable AI the is far more advanced then asking it to solve a 2 + 2 question.

The article worded it perfectly; friction-less "efforts"

jackbucks• 1 month ago

Excuse me for asking, doesnt this make it easier to run a malware bot farm and disappear without a trace?

dools• 1 month ago

The idea of giving a non deterministic automated process direct deployment control is fucking madness to me. That’s why I don’t get the obsession with MCP. Deployment can be scripted. It doesn’t need an LLM, it is a completely deterministic process and you want it to run identically every single time.

The right model for agentic API usage is having LLMs write scripts that use APIs. Connecting agents to MCPs and telling them to go and do stuff over and over not only wastes money but invites catastrophe.

taosu_la• 1 month ago

I am a bit worried about being misused by black and gray industries, which could then lead to Cloudflare or the US Congress imposing various restrictions, resulting in the functionality of my ordinary Cloudflare account being limited.

FergusArgyll• 1 month ago

https://codex-cloudflare-skill-site.aspiring-sheep.workers.d...

Temporary Cloudflare accounts for AI agents

Related Stories