SWE-agent is an MIT-licensed autonomous software-engineering agent from Princeton and Stanford researchers that takes GitHub issues or coding tasks and attempts to solve them with a language model of your choice. It remains one of the canonical open-source references for agent-computer-interface design, repository navigation, file editing, test execution, and patch generation.
The important current-status caveat is mini-swe-agent. The SWE-agent README now says most current development effort is on mini-swe-agent, which has superseded SWE-agent, matches its performance while being much simpler, and is the general recommendation going forward. SWE-agent itself is still public, not archived, and useful as a deeper research/reference implementation rather than a polished hosted product.
SWE-agent runs locally or in your own CI environment with bring-your-own LLM API keys. It is strongest for researchers, open-source maintainers, and advanced engineering teams that want to study or customize an autonomous coding loop. Teams seeking production fleet management, review dashboards, audit trails, or turnkey cost controls should compare mini-swe-agent, OpenHands, or managed commercial coding-agent platforms before standardizing on SWE-agent.
