SWE-agent
SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive co
What is SWE-agent?
SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive coding challenges. [NeurIPS 2024]
Overview of SWE-agent
What it is
SWE-agent is an open-source tool that takes a GitHub issue and attempts to automatically fix it using a language model of your choice (e.g., GPT-4o or Claude Sonnet 4). It can also be employed for offensive cybersecurity (capture the flag) challenges or competitive coding challenges. The project is built and maintained by researchers from Princeton University and Stanford University, and was presented at NeurIPS 2024. Note: The current development focus has shifted to mini-swe-agent, which matches SWE-agent's performance while being simpler. The general recommendation is to use mini-swe-agent going forward.
Key features
- State of the art on SWE-bench among open-source projects - Free-flowing and generalizable, leaving maximal agency to the language model - Configurable and fully documented, governed by a single YAML file - Made for research: simple and hackable by design - Includes a mode for offensive cybersecurity (EnIGMA) that achieves state-of-the-art results on multiple cybersecurity benchmarks
Use cases
- Automatically fixing bugs in real GitHub repositories - Finding cybersecurity vulnerabilities (capture the flag challenges)
- Performing custom tasks that require autonomous tool use by a language model
Getting started
See the repository README for installation and quick-start instructions.
License and pricing
MIT license. Open source.
Based on the project's public README and repository metadata.
