SWE-agent
Autonomous coding agent that fixes real GitHub issues — now succeeded by mini-swe-agent.
About
SWE-agent lets a language model autonomously use tools (shell, file editor, browser) to fix issues in real GitHub repositories, find vulnerabilities, or solve coding tasks — configured by a single YAML file and built for research hackability. It held state-of-the-art on SWE-bench among open-source projects (SWE-agent 1.0 + Claude 3.7 topped the verified board in early 2025). 20.5k stars, MIT. Important: the README states development effort has moved to mini-swe-agent, which 'has superseded SWE-agent' at 65% on SWE-bench Verified in 100 lines of Python — use mini-swe-agent for new work. Now under the SWE-agent org (moved from SWE-bench; old URLs 404).
What it can do
- **Autonomous repo work**: explore, edit, test, and iterate on real codebases
- **YAML configuration**: models, tools, and agent behavior in one file
- **Research-first**: simple, hackable design for agent experiments
- **SWE-bench pedigree**: built by the benchmark's own team (Princeton/Stanford)
- **Successor path**: mini-swe-agent for simpler, maintained development
Setup
```bash
pip install sweagent # or run via Docker (recommended)
# set MODEL_NAME + API key, then:
sweagent run --agent.model.name gpt-4o --problem_statement.text "Fix the failing test"
```
Or try the successor: github.com/SWE-agent/mini-swe-agent. Docs: swe-agent.com.
Prerequisites
Python 3.11+; Docker recommended (agent runs code in containers). An LLM API key (GPT-4o, Claude, etc.).
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: 20,537 GitHub stars (log-scaled vs catalog peers); blended with ~138 monthly package downloads; last push 7d ago; 0 releases/12mo; 158 open issues. Manual review: research reference for coding agents, SOTA on SWE-bench (OSS); 20.5k stars; BUT README: dev moved to mini-swe-agent which 'superseded SWE-agent'; 0 releases/12mo. License: MIT.
★ 20,537 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.8/10
- Maintenance · recency, releases, issue load
- 7.6/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · 20,537 GitHub stars (log-scaled vs catalog peers)
- · blended with ~138 monthly package downloads
- · last push 7d ago; 0 releases/12mo; 158 open issues
- · manual review: research reference for coding agents, SOTA on SWE-bench (OSS); 20.5k stars; BUT README: dev moved to mini-swe-agent which 'superseded SWE-agent'; 0 releases/12mo
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.