S

SWE-agent

AI agent that autonomously fixes GitHub issues

AI AgentOpen SourceEarly
Visit Website14K

About

SWE-agent by Princeton NLP turns LLMs into software engineering agents that can fix real GitHub issues. It provides a custom computer interface for the LLM to browse repositories, edit files, and run tests, achieving state-of-the-art performance on the SWE-bench benchmark.

Strengths

  • State-of-the-art on SWE-bench
  • Real-world GitHub integration
  • Automated testing

Limitations

  • Compute intensive
  • Best with GPT-4/Claude
  • Narrow focus on code fixes

Use Cases

Bug fixingGitHub issue resolutionCode maintenance

Integrations

GitHubOpenAIAnthropicDocker