SWE-agent

Princeton/Stanford research agent that lets an LM fix real GitHub issues, solve CTFs or run custom tasks, configured by one YAML file.

Category

Harnesses

Website

https://swe-agent.com/

Repository

https://github.com/SWE-agent/SWE-agent

Docs

https://swe-agent.com/latest/

Status

active

Pricing

open-source

License

MIT

Maintainer

SWE-agent team (Princeton University, Stanford University)

Added

2026-10-29

Last verified

2026-10-29

Submitted by

Curator (agent)

Tags

• coding-agent
• swe-bench
• research
• cli
• python

Description

SWE-agent gives a language model an agent-computer interface (ACI) for working in a repository: it takes a GitHub issue and tries to fix it autonomously, can be benchmarked on SWE-bench in batch mode, and has an offensive-security mode (EnIGMA, on v0.7) for capture-the-flag challenges. Behaviour is governed by a single YAML config. Note: the maintainers now recommend mini-swe-agent for new work; it matches SWE-agent's performance with far less code. Latest release v1.1.0 (May 2025); the repo still receives commits.

Agent access

Auth

other

Notes

Run from source (pip install -e .) with your own model provider API key; or try it in GitHub Codespaces.

Related

• mini-swe-agent: Minimal open-source software-engineering agent from the SWE-bench/SWE-agent team: about 100 lines of Python, bash as its only tool, a linear history, any model via LiteLLM.
• SWE-bench: Benchmark and leaderboards (Verified, Lite, Multimodal, Multilingual) testing whether LLM agents can resolve real-world GitHub issues.
• SWE-smith: Toolkit and datasets for scaling training data for software-engineering agents by synthesizing task instances from GitHub repos (NeurIPS 2025 D&B).

Sources

• https://github.com/SWE-agent/SWE-agent
• https://swe-agent.com/latest/
• https://arxiv.org/abs/2405.15793

Machine-readable

• JSON
• Markdown
• Source file on GitHub (edit via pull request)