Hey all! Jay here from Codegen. We're incredibly excited to launch our generalist SWE agent today for general availability! 🎉
2 years ago, we set out to build an agentic developer. With Claude 3.7, it's now possible like never before - and it's absolutely killer. It's essentially Claude Code tightly integrated with Github, Slack, Linear and more - the modern developer toolkit.
This is not a contrived demo. We're constantly merging PRs in enterprise repos. It's often faster to kick off Codegen in a Slack thread than to open VSCode, push up a branch, etc. We've found it *especially* useful in scenarios where multiple humans are collaborating real-time. Fastest way to push collaborative dev initiatives forward.
I've been at the bleeding edge of code generation for years now and it's incredible to me how far we've come. Claude 3.7 is shockingly resilient, resourceful, and will go for *hundreds* of iterations even in the face of ambiguity, broken tools, etc. It just finds a way.
Accelerate your development today! Install at https://codegen.sh/install
And as always, we love feedback from the community. I'm here to answer questions and chat ideas with anyone who's curious. Thanks!
Oh my gosh, I can't believe this is possible in Slack. I understand modifying the code on Github, but do I have to check manually that it actually works without errors?
@kay_arkain great point - if you are doing something that requires extensive manual debugging, then I would recommend a more in-the-loop development process, for example Windsurf/Cursor/Cody. If, however, the bugs are something that would be caught by your tests/linter/CI, then Codegen is capable of reacting to your tests!a No need to pull down the changes.
Where Codegen really shines through is in scenarios like small bug fixes, replicating boilerplate, asking questions about your code, etc., and you want the result fast
Congratulations on the launch. Just watched the demo video. Codegen is indeed like having a software engineer who never sleeps!
How does Codegen handle large monolithic repositories compared to microservices architectures? Are there specific optimizations to ensure efficient performance across diverse codebase structures?
@faizanjan_ Thanks! We've implemented many optimizations to make it work well with large repositories, including optimized search functionality and more. That being said, in our experience the current generation of foundation models (Claude 3.7 etc.) are actually incredibly skilled at navigating large repositories even given a limited toolset. Try it yourself and I think you'll find it capable of doing really valuable work even in large repos or scenarios with hundreds of microservices!
Mira
Oh my gosh, I can't believe this is possible in Slack. I understand modifying the code on Github, but do I have to check manually that it actually works without errors?
Mira
@kay_arkain great point - if you are doing something that requires extensive manual debugging, then I would recommend a more in-the-loop development process, for example Windsurf/Cursor/Cody. If, however, the bugs are something that would be caught by your tests/linter/CI, then Codegen is capable of reacting to your tests!a No need to pull down the changes.
Where Codegen really shines through is in scenarios like small bug fixes, replicating boilerplate, asking questions about your code, etc., and you want the result fast
Congratulations on the launch. Just watched the demo video. Codegen is indeed like having a software engineer who never sleeps!
How does Codegen handle large monolithic repositories compared to microservices architectures? Are there specific optimizations to ensure efficient performance across diverse codebase structures?
Mira
@faizanjan_ Thanks! We've implemented many optimizations to make it work well with large repositories, including optimized search functionality and more. That being said, in our experience the current generation of foundation models (Claude 3.7 etc.) are actually incredibly skilled at navigating large repositories even given a limited toolset. Try it yourself and I think you'll find it capable of doing really valuable work even in large repos or scenarios with hundreds of microservices!