#3737th of 37 in Coding agents and assistants
swe-agent
Research agent that lets an LLM fix GitHub issues using tools
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 20.6k
- License
- MIT
- Last commit
- Jul 2026
- Last release
- May 2025
- Language
- Python
18,647 on 24 Mar 202620,552 on 13 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
SWE-agent lets a language model such as GPT-4o or Claude Sonnet 4 autonomously use tools to fix issues in real GitHub repositories, find cybersecurity vulnerabilities, or run custom tasks. Behavior is governed by a single yaml file, and it can be run in batch mode for SWE-bench. The README says the maintainers now focus on mini-swe-agent, which supersedes it, and recommend using that instead.
Who it is for: Researchers and engineers benchmarking LLM agents on SWE-bench
Strengths
- Single yaml file configures agent behavior; documentation site covers install, usage and batch mode
- Batch mode for benchmarking on SWE-bench
- Works with the model of your choice, e.g. GPT-4o or Claude Sonnet 4
- MIT license, built by Princeton and Stanford researchers with an accompanying paper
Weaknesses
- README says mini-swe-agent supersedes it and recommends using that instead
- Last release was 2025-05-22, well behind recent commits
- EnIGMA cybersecurity mode requires SWE-agent 0.7 until it is updated for 1.0
- Built for research and benchmarks rather than daily IDE-style assistance
What it needs
- no GPU
- Models: GPT-4o, Claude Sonnet 4, Claude 3.7
Also in Coding agents and assistants
See all 37| Rank | Project | Score |
|---|---|---|
| 1 | claw-codeRust CLI coding agent harness using Anthropic and OpenAI-compatible APIs | 83 out of 100 |
| 2 | opencodeTerminal coding agent with build and plan modes, plus a desktop app | 79 out of 100 |
| 3 | oh-my-openagentTerminal coding agent that fans work out across many models | 78 out of 100 |
| #4 | orcaDesktop app that runs CLI coding agents in parallel git worktrees | 77 out of 100 |
| #5 | openinterpreterTerminal coding agent, a Codex fork tuned for low-cost models | 77 out of 100 |