SWE-agent
Princeton NLPResearch-driven autonomous coding agent from Princeton NLP that takes a GitHub issue and attempts to automatically fix it using LLMs. Also supports offensive cybersecurity and competitive coding challenges. Published at NeurIPS 2024 and served as an early benchmark-setter for the SWE-bench leaderboard. 19.3K stars with 92 contributors; development pace has slowed in 2026 (16 commits in 90 days) but the project remains influential as a reference architecture.
SWE-agent (Princeton/Stanford), latest release v1.1.0 (May 22 2025); GitHub SWE-agent/SWE-agent 19.4k stars. Research agent that lets an LLM autonomously fix GitHub issues. Last tagged release >12mo old at scoring time (foundational research tool, repo still referenced).
Openness
5 high confidence- license
- MIT(OSI)
- source
- public
- core-gated
- ungated
Verified MIT OSI, full source public; academic project with no proprietary tier. License/stars/release date confirmed via primary GitHub source.
- https://github.com/SWE-agent/SWE-agent/blob/main/LICENSE recorded 2026-06-04
MIT License text
- https://github.com/SWE-agent/SWE-agent recorded 2026-06-04
19.4k stars, v1.1.0 May 22 2025, Princeton/Stanford (verified)
Adoption
2 low confidencePrimarily a research/benchmark agent; 19.4k stars in the research community. No published download/user volume; capped per stars-fallback rule. Adoption is research-circle, not production end-user scale.
- https://github.com/SWE-agent/SWE-agent recorded 2026-06-04
19.4k stars, 2.1k forks (research adoption)
Capability
3 medium confidenceOriginator of the SWE-bench agent scaffold; mid-tier on the 2026 leaderboard (43.2%), above Aider but below Cline/OpenHands. NOTE: figure from aggregator (awesomeagents.ai), not primary.
- https://awesomeagents.ai/leaderboards/swe-bench-coding-agent-leaderboard/ recorded 2026-06-04
SWE-agent v1 SWE-bench Verified 43.2% (aggregator)
Unchanged since 2026-07-30 (last edited, not re-checked)