Autonomous Tool Use
Enables language models to autonomously use available tools to fix issues or perform tasks in GitHub repositories.
AgentsSWE-Agent
By Princeton University and Stanford University researchers
SWE-agent enables language models to autonomously fix GitHub issues, detect vulnerabilities, and perform custom coding tasks with high flexibility and configura
SWE-agent is an autonomous agent framework that allows language models such as GPT-4o or Claude Sonnet 4 to use tools and interact with real GitHub repositories to fix bugs, find cybersecurity vulnerabilities, or handle any custom tasks configured by the user. It is designed for generalizable and free-flowing natural language decision making, leaving maximal agency to the underlying language model.
The project is state of the art on SWE-bench benchmarks among open-source projects and emphasizes configurability through a single YAML configuration file, making it suitable for research and experimental modifications. While SWE-agent itself is now in maintenance mode, its successor, mini-swe-agent, offers comparable performance with a simpler and more flexible design.
Maintained by researchers from Princeton University and Stanford University, SWE-agent supports a variety of language models and provides extensive documentation for installation, configuration, and customization. It also includes support for vision-capable models to process images from GitHub issues, expanding its multimodal capabilities.
Enables language models to autonomously use available tools to fix issues or perform tasks in GitHub repositories.
Entire behavior governed by a single YAML configuration file allowing full customization and extension.
Achieves state-of-the-art results on SWE-bench among open-source projects.
Processes images from GitHub issues using vision-enabled AI models.
Designed to be simple, hackable, and adaptable for research purposes.
Supports various language models such as GPT-4o and Claude Sonnet 4.
Use SWE-agent to autonomously identify and fix bugs in your GitHub repositories.
Detect and address security vulnerabilities in codebases automatically.
Configure SWE-agent to perform custom development or research tasks with language models.
Process and analyze GitHub issue images using vision-capable AI.