Description
SWE-agent is an open-source AI agent that uses language models such as GPT-4o to solve GitHub issues, fix bugs and implement features in a code repository. You configure its behavior with YAML files, pick your model and tools, and run it on a single problem, in batch mode or in a competition setup. Its documentation covers installation from source or in a browser, model keys, the command line interface, configuration templates, tools, environments and a trajectory inspector to see each step the agent took. The project's site notes that the team now recommends mini-swe-agent, a simpler project with similar performance that has superseded SWE-agent for most new development. The code and documentation are on GitHub, and the project has a contribution guide and API reference for developers. The site also lists a changelog and tutorials on the basics.…
Preview
Highlights
- Solves GitHub issues with LLMs
- YAML-configured agent
- Batch mode and trajectory inspector
- Custom tools and environments
- Open source on GitHub
Ratings & Reviews
No ratings yet
Be the first to rate SWE-agent.
People also visit
SWE-agent FAQ
What is SWE-agent?
An open-source AI agent that uses language models to solve GitHub issues and implement changes in a repository.
Is it free?
Yes. The code is on GitHub. You bring your own model keys.
Is it still the recommended tool?
The site recommends mini-swe-agent instead, which it says matches SWE-agent's performance and is simpler.
How do I configure it?
With YAML config files for the agent, models, tools and environments.
Can I see what the agent did?
Yes. A trajectory inspector shows each step.
What are the best SWE-agent alternatives?
Popular alternatives to SWE-agent include DeepSeek-Reasonix, Easy Agent and RepoMaster. See all SWE-agent alternatives

