They "gift you" a free standard plan if you have above a certain (non-transparent) level of stars, I don't think you can even disable your "subscription" if you get it for free.
Pleasure! Yeah it's just the standard runner - the only thing we've built is a duplicate control-pane, which is just HTTP. Since I have the standard runners source code it's pretty trivial for agents to know what to implement on the other end.
I built this because I treat CI as the last line of defense. Agents also need validation. They should use CI, and they shouldn't bother you unless everything is green!
GH Actions is usually in the top-5 expenses for dev-teams. Add agents to that mix? It'll easily double. It's the wrong tool for the right job: Slow boot, slow cache, retrieving logs is token expensive for agents, the list goes on...
So I built a tool with one amazing feature: live-reload for failures. Agent-CI is a local CI runner.
I tweaked the control pane and mounts to provide 0ms caching, insanely fast boots. When a step fails it pauses, provides the agent with the failure, and waits for the agent to fix and retry just that step.
It uses the standard GH Actions image (via Docker), but emulates the control pane via a local HTTP server. You don't have to change any of your existing GH workflows.
The big focus will be ensuring that you're not stuck on our implementation of RSC (we wrote a custom vite plugin); so we'll be moving over to the version from the Vite team.
It's not that we think our implementation is worse, it's just that we don't want you to be locked into our framework.
We will also launch something like ActiveRecord and a Convex like experience.
Hey everyone, Peter here. I'm one of the humans behind RedwoodSDK.
This is the second time I’ve built this framework.
The first version was a collaboration with Tom Preston-Werner. When Tom asked me to become the steward of RedwoodJS, I realized we had a massive opportunity to take what we’d started and make it bulletproof.
I brought my experience from my own startup to the table, specifically the scars , to ensure this version allowed you to iterate rapidly.
Congrats on the launch! While the focus is on humans, I found many of the principles here to be very relevant for building with AI.
With LLMs able to generate the "magic" with ease, I can see the role of frameworks shifting a bit. Instead of hiding complexity, it's surely more beneficial to expose the primitives clearly enough that both you and an LLM can follow the request lifecycle.
The bet on Cloudflare is pretty bold, but it does make sense given how unique their products like DO and Workers are. I suppose there's not much of an escape hatch if you ever needed one?
> With LLMs able to generate the "magic" with ease, I can see the role of frameworks shifting a bit. Instead of hiding complexity, it's surely more beneficial to expose the primitives clearly enough that both you and an LLM can follow the request lifecycle.
100%.
I kinda feel like we've fooled ourselves into creating these domain specific languages for frameworks, where we glue things together using magic, and just dealing with the features of the language, the browser and the network make a lot of sense in the AI world. And our framework is actually much less complicated than I think a lot of them are...
> I suppose there's not much of an escape hatch if you ever needed one?
I of course don't want to be vendor locked in, so I've started working on something called open durable objects, and I have a concept for something called open bindings.
these are the two core technical solutions that the framework really needs from Cloudflare. This does not include durability in the same way you get it from Cloudflare: it will run on a single machine.
reply