SWE-agent

SWE-agent is the open-source autonomous coding agent that automatically fixes GitHub issues by reading the issue description, exploring the codebase, making code changes, and submitting pull requests — using whatever LLM the user configures. Developed by researchers at Princeton University, Carnegie Mellon, and other academic institutions and published as a NeurIPS 2024 paper, SWE-agent has become one of the most academically-credentialed autonomous coding agents and a defining benchmark in the SWE-bench (Software Engineering benchmark) leaderboard category. Pricing follows the open-source BYOK (Bring Your Own Key) model: SWE-agent itself is free under MIT license, with users paying only for LLM API usage through their chosen provider. The platform supports Claude, OpenAI GPT, Google Gemini, and any OpenAI-compatible endpoint. Typical SWE-agent runs cost $1-$10 per issue depending on issue complexity and chosen model, materially cheaper than commercial autonomous coding tools (Devin, Claude Code) at moderate use. SWE-agent's differentiation versus Devin, Claude Code, and Cursor Background Agent is the academic credibility combined with research transparency: rather than being a commercial product with proprietary methods, SWE-agent is research-grade open-source code with published methodology, full reproducibility, and benchmark transparency on SWE-bench. This matters for academic researchers, security auditors, and organizations that want to understand exactly how autonomous coding agents work before deploying them. SWE-agent has been employed for offensive cybersecurity research and competitive coding challenges in addition to standard issue-fixing use cases. The platform is open-source on GitHub under the SWE-agent organization with active community contribution from academic institutions and individual developers. The 19,100+ stars and 2,100+ forks reflect strong adoption in the academic and research communities. SWE-agent operates as a research tool rather than a productized commercial offering, with no commercial support, SLA, or compliance certifications — appropriate for the project's positioning as research infrastructure rather than enterprise software.