[P] SWE-agent: an open source coding agent that achieves 12.29% on SWE-bench

We just made SWE-agent public, it's an open source agent that can turn any GitHub issue into a pull request, achieving 12.29% on SWE-bench (the same benchmark that Devin used).

​

[https://github.com/princeton-nlp/swe-agent](https://github.com/princeton-nlp/swe-agent)

​

We've been working on this for the past 6 months. Building agents that work well is much harder than it seems- our repo has an overview of what we learned and discovered. We'll have a preprint soon.


We'll hang out in this thread if you have any …

agent agents benchmark building coding devin github issue machinelearning open source public request swe work