Radicle: Sovereign {code forge} built on Git

by KolmogorovComp • 2 months ago

274 93 2 months ago

minraws2• 2 months ago

The more I have been using git and building my own tooling and services around it for usage, I have figured out that something like radicle feels like the right/better solution, definitely better than what github is atm.

There are rough edges and the seeding thing is a bit mehhh. And honestly there are a bunch of things I would do differently but I like the spirit of things.

Not sure where the authors of the project stand, but it's fun to see them make progress.

endiangroup8• 2 months ago

AD: We're definitely interested in hearing about the rough edges and opinions! Feel free to post on our Zulip [^1] if you haven't already :).

[^1]: https://radicle.zulipchat.com/

adastra226• 2 months ago

What are the things you would do differently?

minraws2• 2 months ago

Prioritize discoverability, and self hosting on my own domain, with more customization for the protocol and ui/ux aspects.

Basically allow people to deploy something closer to gitea on a webserver initially, which has all the basic protocol features and sharing aspects implemented.

Currently the focus seems to have been very much on just the seeding & p2p part which honestly is good but not very useful for me as a developer/user who wants to work on a collaborative project.

Prioritizing P2P makes sense, but the UX/DX of radicle is not intuitive to put it mildly.

I went to the explorer radicle.network and my own as well and I still had no idea how to create an issue or submit a patch or login to an account or anything of that sort.

I opened the cli and figured out how to do all that, but then I created another node on my web server and now I have no idea how to connect to it, then I figured out how to do that, now how do I share my node without sharing it publicly, figured that out.

Cool now we are at the starting point, so every time I need to work on a project i will have to clone it, but why can't I create stuff in webui.

Maybe there is a setting I can't seem to find it yet.

But the jumping through hoops is mindbending. For a solution that could replace github for me and my friends who maintain a few public projects that wants small contributions from large number of folks as we try and maintain the local "tech" clubs.

Getting low information folks to use within 10 mins, is the bar that I have set for any platform I can use to replace github.

Radicle isn't that yet! anyways.

But I like the direction, putting contributors, access list, issues and prs in the repo is a brilliant idea, maybe put agent plans in it too, agent sessions as well maybe...

I would go as far as to say, something like a radicle network is better than every other alternative, but the UX is just not there yet.

josh-sematic• 2 months ago

Love to see it. Unlike tangled.org this is local-first and has a solid story around private repos. I’m bullish on distributed forges in general, but I’m all for experimentation in figuring out exactly what that looks like.

vaylian• 2 months ago

I noticed that they moved to a different domain last month: https://radicle.dev/2026/04/23/domain-move

satvikpendem• 2 months ago

.xyz and similar domains are often blocked by ISPs due to higher than usual rates of spam from domains using that TLD.

deaton• 2 months ago

Makes sense. Usually if you want a 99 cent registration, .xyz is whats available.

einpoklum• 2 months ago

I like this idea a lot! I need to find people to try it with, though, which is not so easy with GH being so popular :-(

Some nitpicks:

* What is with the forced serif font on the website?

* Does this support other version control systems? Like mercurial, SVN, pijul, etc.?

endiangroup8• 2 months ago

AD: The website is currently being redesigned, I suspect we'll have it up in the next few weeks.

No it doesn't currently support other VCS's but we have planned for that possibility in future!

esafak4• 2 months ago

VCS agnosticism would be a selling point.

backscratches• 2 months ago

I read that it works well with jujutsu

gsaslis6• 2 months ago

yup ! https://radicle.dev/2025/08/14/jujutsu-with-radicle

esafak4• 2 months ago

How well does it hold up under load? What are the CI and PR stories?

h1watt2• 2 months ago

Their CI take is early but promising: https://radicle-ci.liw.fi/

jayd162• 2 months ago

This just describes a watcher service that kicks jobs off on an external CI system and logs the results? Not much more detail than that.

Gitlab and Github have pages and pages going over the domain language used to configure the job triggers. Jobs can trigger other jobs either in response to completion or as a dependency etc etc.

I would say these radicle-ci designs as they are now are actually quite rudimentary. That's perfectly fine for an early project but at this point I think you have to say that they won't have a CI system ready for quite some time.

h1watt2• 2 months ago

Radicle is really underrated, especially when working with agents. I find it a joy to use for my agentic workflows.

If there's purely an agentic forge one day, it's likely going to be a distributed one, with cryptographic identities and signed artifacts by default.

eightysixfour• 2 months ago

What makes it a joy to use with agents?

k__• 2 months ago

Probably less issues with account management and rate limits.

endiangroup8• 2 months ago

AD: Feel free to post on our Zulip [^1] about your experiences of agentic workflow if you haven't already! Some of the team are interested in developing the agentic workflow experience.

[^1]: https://radicle.zulipchat.com/

chrisweekly• 2 months ago

"AD" ?

xvilka4• 2 months ago

I am surprised you don't use RustChat[1]

[1] https://rustchat.io/

bfrog• 2 months ago

radicle is awesome and Just Works from what I've tried of it

pessimizer4• 2 months ago

I'd like to see radicle replace crates.io. I can't get over Rust's dependency on github/Microsoft, and I can't get over the lack of namespacing.

All you would need is cargo compatibility, and a trusted namespace that kept up with the metadata of the current contents of crates.io, right?

edit: I really, really like rust, and love basically all of their choices about the language, but I can't stand the feeling that I'm being tricked into an ecosystem dependent on one of the worst behaved companies in the world, and I can't stand that a lot of rust projects smell like GPL-washing.

That being said, git is GPL and radicle is MIT, so it feels like the same thing, but Github also ain't git. I prefer MIT to MS; if radicle gets important enough and decides to rubpull, there will inevitably be a Free fork anyway.

Pay083• 2 months ago

Does radicle have some way of storing binaries outside of the source tree? I know cargo compiles from source by default, but AFAIK it can (and does) download binaries as well.

endiangroup8• 2 months ago

AD: Theres a project `artifact-cob` [^1] that enables some of the functionality you are looking for. You can host artifacts off-network and make 'releases' that point to them.

[^1]: https://radicle.zulipchat.com/#narrow/channel/369274-General...

2color5• 2 months ago

radicle-artifact is currently under active development: https://radicle.network/nodes/iris.radicle.network/rad:z4VYy...

pie_flavor2• 2 months ago

* crates.io is moving away from GitHub-only authentication

* crates.io's attachment to GitHub is a fact about crates.io specifically, not the Cargo crate registry protocol

* Cargo's support for Git repositories is generic across Git and has nothing to do with GitHub specifically

* Radicle offers nothing to a crate registry that a Git remote doesn't

* and none of this has anything to do with the GPL.

It feels like you're just listing off things you like and don't like aesthetically. They have nothing to do with each other structurally.

pessimizer4• 2 months ago

crates.io is moving away from GitHub-only authentication

Crates.io has not moved away from Github-only authentication, and got into the habit of yelling at people who complained about it.

crates.io's attachment to GitHub is a fact about crates.io specifically, not the Cargo crate registry protocol

Is this just trivia you wanted to share? I feel like I covered it in the second sentence of the comment you're replying to.

Cargo's support for Git repositories is generic across Git and has nothing to do with GitHub specifically

I'm looking to compile Rust projects from the semi-standard commonly-used crates. I do not want a Github account.

Radicle offers nothing to a crate registry that a Git remote doesn't

Radicle offers peer-to-peer hosting that does not require a Github account.

none of this has anything to do with the GPL.

Radicle is a project being built in Rust that partially reimplements git. Git is GPL-licensed, Radicle is MIT-licensed.

It feels like you're just listing off things you like and don't like aesthetically.

I am unconcerned about your feelings. What I was saying is that I would like a peer-to-peer hosted, namespaced code repository that mirrors (or replaces) crates.io, and I do not want a github account to be necessary to use it.

They have nothing to do with each other structurally.

I have no idea what "they" is referring to in this sentence.

crabbone2• 2 months ago

I tried to understand what this does...

What is Radicle? How is it different from Git/GitHub?
Radicle is a peer-to-peer code collaboration platform (“forge”) built on Git. Unlike centralized platforms like GitHub, there is no single entity controlling the network or user data. Repositories are replicated across peers in a decentralized manner. Radicle is an alternative for people and organizations who want full control of their data and user experience, without compromising on the social aspects of collaboration platforms.

(Quote from their FAQ).

This isn't even trying to answer the titular question... None of them, actually.

So, what is Radicle? A platform built on Git? What does this mean? A platform for what? What is it for?

Why Git/GitHub are used as if they were the same category of things? There's not even an attempt at answering the "how is this different from Git?" question. What does it offer that Git doesn't? Wtf is "forge"?

Radicle is an alternative... to what? I believe I have full control of my data in my Git repository... why do I need an alternative with even more control? How will I have even more control?

* * *

Maybe whatever this software does is actually useful or even good, but the documentation can't be worse.

pessimizer4• 2 months ago

So, what is Radicle?
Radicle is a peer-to-peer code collaboration platform (“forge”) built on Git.

-----

Why Git/GitHub are used as if they were the same category of things?

They are not. Github is a centralized collaboration platform built on git, and radicle is a peer-to-peer collaboration platform built on git.

-----

Wtf is "forge"?

A word some people started using for the class of Github/Bitbucket(RIP) or even Fossil-type things, as FOSS alternatives began to multiply.

-----

Radicle is an alternative... to what?

To Github, or other "forges."

47282847• 2 months ago

The first I remember using it was SourceForge (1999).

crabbone2• 2 months ago

To Github, or other "forges."

Then why do they bring Git into the picture? They are not comparing themselves to Git...

Radicle is a peer-to-peer code collaboration platform (“forge”) built on Git.

This is a word salad that means nothing... more than 99% of moderns software is built on Git in one way or another. Anything that is designed to be used by more than a single user could be arguably called a "collaboration platform". This description completely fails to describe anything useful about the program they are trying to describe. For instance, Git is a peer-to-peer code collaboration platform built on Git. And the same can be said about a huge number of programs that share very little in terms of purpose or application.

When someone writes a definition, s.a. you'd find in encyclopedia, the rule is that it has to (a) link to the broader category of things (b) specify in what way the subject is (mostly) unique in the category (a). If you give only (a), then the reader walks away wondering how is subject different from anything else in (a). If you only give (b), then the reader needs to guess (a), and if they fail, they may misattribute or simply abandon efforts to understand the subject.

This attempt at "definition" is the textbook example of forgetting the (b). It's something that a 10-12 y.o. could come up with... this is not what an adult should strive for.

They are not. Github is a centralized collaboration platform built on git, and radicle is a peer-to-peer collaboration platform built on git.

You didn't understand the question.

ghostpepper• 2 months ago

I agree it's not immediately clear how it works, although I think I understand the role it's intending to fill.

If you're not familiar with the distinction between git and github it could be even more confusing.

As soon as I hear decentralized I have lots of questions about the underlying protocols. Their protocol page helps a little but also uses terms I'm not familiar with like "gossip protocol".

It would be nice for there to be a page that motivates the project a bit more, ie. explaining the technical problems they are attempting to solve before enumerating the components of the complex system they've built.

adastra226• 2 months ago

Someone looked at git - a distributed version control system that already works, that has been working since 2005, that Linus Torvalds wrote in a fit of pique and spite, and that currently hosts approximately all of the world's source code - and said, "This is good, but what if we added BitTorrent?" And then, presumably after consuming substances that I cannot legally inquire about, they continued: "And what if we added the Bitcoin peer-to-peer protocol?" And then, reaching a crescendo of architectural ambition that would make Icarus say "maybe dial it back a little," they concluded: "And we should DEFINITELY add blockchain identity."

Notice how each additional sentence makes the previous sentence worse, like a turducken of solutions looking for problems, or a nesting doll where every layer is a different kind of sadness. This, I presume, is what happens when you have a hammer and a screwdriver and a chainsaw, and you decide that every problem would be better solved by using all three simultaneously while riding a unicycle.

But seeing as it already does exist, it's pretty awesome.

xvilka4• 2 months ago

I wish they would make local-only deployment easier. For example, lets take 3 machines and try to setup Radicle to work only on those, without joining the common Radicle network. Like on-premises GitLab, but decentralized, without the need of the server. It requires quite some serious scripting and usecase not covered in the documentation.

endiangroup8• 2 months ago

AD: We've just been discussing this! There's an open RIP [^1] that we will be supplying feedback for soon.

[^1]: https://radicle.zulipchat.com/#narrow/channel/369876-RIPs/to...

heresie-dabord• 2 months ago

A containerised deployment would (understatement warning!) be useful.

gsaslis6• 2 months ago

This is a good fit primarily if you want to run a Radicle node that only seeds repos you tell it to. If you want to write code which you publish on Radicle, you need the tool that signs all your work with your private key (of your Radicle identity) - i.e. the `rad` CLI - and running that in a container isn't very useful. (e.g. think about how you'd replace `rad clone`). Having said that, here's a container image I maintain, in case it helps: https://quay.io/repository/radicle_garden/radicle-node

vanous• 2 months ago

Same here... I have been looking at all the self hosting options for a while but swapping one centralized system for another isn't what I would prefer. Running/using local and community gitea/forgejo has been good, but I hope decentralized radicle could be the solution to more independence.

survirtual• 2 months ago

I patched my local radicle to remove the default seeds, and I can put my own seeds as default. It is pretty easy to script some commands that auto add your local nodes when init a repo.

I also added some network rules to block non-local network access to radicle. Not needed but I really wanted it to work only on my lan.

Works great. I setup several skills for agent harnesses and they use radicle + jj + git perfectly. It is fun watching issues pop up and monitoring progress via the flow.

I am building more tools around this workflow because it is so effective. Radicle acts as the long term project memory bank and management. I can write issues and they can automatically be picked up.

What I'm adding is making the issues more searchable and an agent proxy that integrates radicle into calls. among other things.

This is all pretty straight forward to do, I really recommend it.

incompleteCode• 2 months ago

Congrats!

Maybe I'm not the target audience for this, so pardon my ignorance when I ask what problem does this solve? Centralization and censorship?

gsaslis6• 2 months ago

In a nutshell, I summarise it like this: I see Open Source Software is a public good. As a public good, it doesn't belong in any proprietary platform, nor should orgs of any kind be in a position to gate keep it. We should rather host it on a public, peer-to-peer network that everyone can have access to.

toyg• 2 months ago

Centralization is not a small problem to solve, and this looks like it might have done just that. More than censorship, I think the worst problem with github is long-term viability / reliability: they could pull the plug tomorrow and most people would likely lose everything but the raw code.

figbert3• 2 months ago

I discovered Radicle back in 2020 (when their website looked incredible: https://web.archive.org/web/20201201030505/https://radicle.x...). I bounced off of it, in part due to being unable to effectively delete repositories. They used to have an FAQ about that—looks like it's gone now, though the public-private repository area is much more fleshed out (you can make a repo private, in which case no new updates will be publicized but the history will still exist). In truth, it's just profoundly difficult to effectively "delete" things in a decentralized system (see: Matrix, BitTorrent, et. al.). But definitely something to consider; people accidentally upload secrets, and want to have some recourse when that happens.

Still, time has passed and I have become more interested in GitHub alternatives (https://figbert.com/posts/ideating-tragit/). Will likely end up moving to Tangled. But first I need to add support over there for pushing over HTTPS...

lukeck• 2 months ago

It’s a fair point about the difficulty in deleting data in a decentralised system but that’s also true in more centralised systems like GitHub or any other website. Once some data is out there, you have no control over whether it can be removed. Other parties could have copied it and may re-share it. “All” removing data from a centralised system does is slow down this spread, sometimes to the point that it is effectively deleted but that can’t be guaranteed.

The best response I’m aware of is to invalidate any secrets that have been accidentally shared. Sometimes that is easier said than done though. And of course there are plenty of other reasons someone might want to delete data.

figbert3• 2 months ago

Well said, in particular the last sentence: IIRC at the time I was seeking to delete a repo that I had made accidentally so as to clean up my profile.

nine_k• 2 months ago

In general, whatever has been made public, is hard to unmake public. There can always remain a copy.

It is acutely true for peer-to-peer distributed systems like Radicle, but is still true for the more centralized GitHub, and for the Web in general. If you want to be able to walk something back, better don't publush it.

Publishing a token or an ssh key should not be a big deal, such a token or key should be immediately revoked, which is as good as deleted. The problem occurs when the fact is not noticed immediately. A git hook can help avoid such mishaps.

pksunkara• 2 months ago

I would love to know the need behind pushing over HTTPS. Is the SSH protocol not good?

figbert3• 2 months ago

It's not a technical issue but a UX one. It would be a pain to set up pushing via SSH to a knot container on my server because I am use SSH for the host already. And specifying a port is ugly. Doing this over the internet negates all these.

2color5• 2 months ago

What are you missing from Radicle today that would make you chose Tangled over it?

deknos• 2 months ago

I have two questions:

1. Does Radicle also work over TOR? 2. Does Radicle support Git LFS and/or Git Annex?

gsaslis6• 2 months ago

1. yes, it's in the user guide [1] 2. Git LFS should work, yes. Care to try and report back ? ;) Open a new topic on the Zulip #Support channel [2] if you run into any issues.

[1] https://radicle.dev/guides/user#4-embracing-the-onion [2] - https://radicle.zulipchat.com/#narrow/channel/369873-Support

2color5• 2 months ago

Regarding 2, we're actively working on radicle-artifact as an alternative to Git LFS with transport agnostic distribution (https/p2p), attestations, and redactions.

https://radicle.network/nodes/iris.radicle.network/rad%3Az4V...

tsuraan• 2 months ago

How do these federated forges deal with spam? If merge requests and issues are federated, does that mean that anybody running a radicle node (or interacting with one) can open issues or merge requests on all the repositories that you've made public? Or is there a whitelist (or something fancier?) to allow interaction only with specified nodes?

gsaslis6• 2 months ago

yes, we currently have just the base building blocks (`rad follow` - allowlist / `rad block` - blocklist) for the fancier things to be built on. When you "seed" a repository, by default you seed it only with "followed" scope, which means you would only see issues and "Patches" (our term for PR/MR) from the repo maintainers + other peers you follow (i.e. in your allowlist).

Meneth• 2 months ago

Unfortunate that they aren't using the AGPL license. It will allow SAAS companies to Embrace, Extend, and Extinguish.

Indeed, this seems to be already planned. https://radicle.dev/faq "Radworks intends to offer services built on top of Radicle."

endiangroup8• 2 months ago

AD: Radicle is protocol driven, my hope as a contributor is it becomes infrastructure.

Zopieux2• 2 months ago

I saw https://radicle.dev/2025/08/14/jujutsu-with-radicle at the bottom of the homepage and assumed it meant Radicle was planning to support a native jj protocol to decentralize jj repositories. That's unfortunately not the case.

Is there a plan for jj repositories in radicle proper? I'm very tired of git shortcomings.

vinnyhaps• 2 months ago

Maintainer of Radicle here.. Currently, we are focused on building on top of Git, and that is very much baked into the protocol design right now.

We will benefit from jj's compatibility with Git for the time being, and seeing how that evolves.

Interoperating with jj's repository would likely be a bigger task and it is still evolving quite quickly itself.

Zopieux2• 2 months ago

Somehow never heard of this before. It's even packaged in NixOS so I'll probably set up an always-on node.

endiangroup8• 2 months ago

AD: Big shout to our community member Defelo for maintaining those! Come join in!

philsnow• 2 months ago

The story for sensitive/proprietary repositories doesn't feel well-fleshed-out yet:

Radicle supports private repositories that are only shared among a trusted set of peers, not the entire network. These are not encrypted at rest but rely on selective replication and are thus completely invisible to the rest of the network.

There's no structural separation of public / private repositories; this is one bug or fat-finger away from a leak.

I was looking into how to get a radicle node running, but connected to only my own devices.

You can't seem to start the radicle node without it automatically connecting to the (generously provided) global shared seed nodes, and you can't examine/change the config without starting the node (at least with rad subcommands).

If you do `rad auth` and then delete the four seed addresses from ~/.radicle/config.json before starting the rad node, it helpfully re-adds them on startup:

  [...]
  INFO  node     Opening node database..                                                                                                                                                                           
  INFO  node     Address book is empty. Adding bootstrap nodes..                                                                                                                                                   
  INFO  node     4 nodes added to address book
  [...]

(It doesn't add the bootstrap nodes if I instead add a seed "<randomDID>@10.254.254.254:8776".)

Is there a guide to using Radicle like one might use Fossil within a small company / within a small group of people (disconnected from the iris/rosa radicle.network seeds)?

2color5• 2 months ago

Radicle contributor here.

When a repository is initialised as private [1], it's encoded in the repository identity document. The only way to change a repository to be public is to update the identity document; in the case of a repository with multiple delegates (repo maintainers in Radicle nomenclature), such a change requires a quorum to be met. So I'd say it's more than a fat-finger away from leaking.

I was looking into how to get a radicle node running, but connected to only my own devices.

There's an open proposal [2] to introduce the concept of network configuration that would help with these kinds of use cases. Moreover, we're working on other ways to improve the collaboration experience for small teams, so stay tuned and thank you for this feedback!

[1] https://radicle.dev/guides/user#initializing-a-private-repos... [2] https://radicle.network/nodes/iris.radicle.network/rad:z3trN...

samuell3• 2 months ago

Somewhat related:

Show HN: Epiq – Distributed Git based issue tracker TUI

https://news.ycombinator.com/item?id=48155570

theptip• 2 months ago

This is cool. I think it fits the Agentic model a lot better; wrangling PATs and repo permissions on GitHub is quite tedious.

How are folks wiring up CI/CD? Seems you want to trigger compute DAGs on the patches you receive, what’s the latest and greatest here?

2color5• 2 months ago

The general idea is using the radicle-ci-broker [1] to hook into repo events, and trigger different adapters based on the specific CI/CD tool you use.

We'll have more docs on this very soon. Also see https://news.ycombinator.com/item?id=48153535 for more info

[1] https://radicle.network/nodes/seed.radicle.garden/rad%3AzBNX...

zfourhrms• 2 months ago

One challenge with distributed systems is that deletion and rollback expectations differ from traditional centralized platforms. Even if a platform removes data, cached or replicated versions can still persist elsewhere.

Radicle: Sovereign {code forge} built on Git

Related Stories