SWE-Agent

Automated Software Engineering Research Project S Voice & Memory

Basic Information

Product Description

SWE-Agent is a system developed by researchers from Princeton University and Stanford University, designed to enable large language model agents to autonomously use computers to solve software engineering tasks. SWE-Agent receives GitHub Issues and attempts to fix them automatically, supporting the selection of different LLM models. Additionally, it can be used for offensive cybersecurity testing or programming competitions.

The core innovation of SWE-Agent is its custom Agent-Computer Interface (ACI), which significantly enhances the agent's ability to create/edit code files, navigate entire repositories, execute tests, and run programs.

Core Features

  • Autonomous resolution of GitHub Issues
  • Custom ACI (Agent-Computer Interface)
  • Code file creation and editing
  • Repository navigation
  • Test execution
  • Multi-LLM support
  • Mini SWE-Agent (a streamlined version with 100 lines of code)
  • Offensive security testing capabilities

Academic Achievements

  • Paper: Yang, J., Jimenez, C. E., et al. (2024). "SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering"
  • Published in NeurIPS 2024
  • SWE-bench pass@1: 12.5%
  • HumanEvalFix pass@1: 87.7%

Business Model

Open-source academic project, completely free. Primarily used for research purposes and academic contributions.

Target Users

  • AI researchers in software engineering
  • Automated bug-fixing scenarios
  • Programming competition participants
  • Security researchers

Competitive Advantages

  • Academically recognized benchmark in automated software engineering
  • Innovative ACI design concept
  • Transparent open-source research code
  • Benchmark performance on SWE-bench
  • Mini version is extremely simple yet performs excellently (74%+ SWE-bench verified)

Market Performance

  • Published in NeurIPS 2024, significant academic influence
  • Widely cited and used on GitHub
  • Advanced the entire field of AI-driven automated software engineering
  • Provided an academic foundation for subsequent commercial products (e.g., Devin)

Relationship with OpenClaw

SWE-Agent represents the academic forefront of AI agents in the software engineering field, and its ACI design concept holds reference value for agent platforms like OpenClaw. Both explore the boundaries of AI agents' ability to autonomously interact with computers.