SWE-agent
Princeton NLPResearch coding agent from Princeton NLP that takes a GitHub issue and attempts to fix it autonomously, and also runs offensive-security and competitive-programming tasks. It was published at NeurIPS 2024, and its agent-computer-interface scaffold became a reference architecture for later agents.
Distributed by clone rather than by package; the PyPI project sees little traffic.
Openness
5 high confidence- license
- MIT(OSI)
- source
- public
- core features withheld
- no
SWE-agent is MIT with its full source public, an academic project with no proprietary tier.
- https://github.com/SWE-agent/SWE-agent recorded 2026-08-13
MIT license on the repo, full agent source public, 20.1k stars, 2.2k forks; no managed tier or enterprise directory
- https://github.com/SWE-agent/SWE-agent/blob/main/LICENSE recorded 2026-08-13
MIT License text, Copyright (c) 2024 John Yang, Carlos E. Jimenez and co-authors
- https://raw.githubusercontent.com/SWE-agent/SWE-agent/main/LICENSE recorded 2026-09-29
The LICENSE file is the MIT License with "Copyright (c) 2024 John Yang, Carlos E. Jimenez" and co-authors, granting the right to "use, copy, modify, merge, publish, distribute, sublicense, and/or sell" the software.
- https://raw.githubusercontent.com/SWE-agent/SWE-agent/main/pyproject.toml recorded 2026-09-29
The package metadata lists "License :: OSI Approved :: MIT License" and points the license at "LICENSE".
- https://raw.githubusercontent.com/SWE-agent/SWE-agent/main/README.md recorded 2026-09-29
The README states the license as "MIT. Check `LICENSE`." and says the project is "built and maintained by researchers from Princeton University and Stanford University"; searched for enterprise, paid tier and license key: none.
Adoption
3 low confidenceSWE-agent publishes no download or user figure, so GitHub stars on its single repository are the only adoption signal available; because a star reflects interest rather than use, that signal alone cannot place it above early adoption. Its adoption is concentrated in the research community rather than production use.
- https://github.com/SWE-agent/SWE-agent recorded 2026-08-13
20.1k stars, 2.2k forks; no download or user count published
Capability
3 medium confidenceOriginator of the SWE-bench agent scaffold, and mid-tier on the current leaderboard at 43.2% SWE-bench Verified (SWE-agent v1 on Claude Sonnet 4.5), above Aider but below Cline and OpenHands. The figure comes from an aggregator leaderboard (awesomeagents.ai) rather than a primary publication.
- https://awesomeagents.ai/leaderboards/swe-bench-coding-agent-leaderboard/ recorded 2026-09-29
Leaderboard row 9 lists "SWE-agent v1", Claude Sonnet 4.5 and "43.2%" on SWE-Bench Verified as open source, and the write-up says "At 43.2% on Verified with Claude Sonnet 4.5, it is not at the frontier".
Verified 2026-09-29, except some scores, still awaiting re-confirmation