SWE-agent
An autonomous AI agent that uses language models to fix issues in GitHub repositories. Simplifies complex software engineering tasks with tool integration and configuration.
SWE-agent is an autonomous AI agent designed to assist with software engineering tasks by leveraging language models to interact with tools and repositories.
Key Features
- Autonomous Issue Resolution: Uses language models (e.g., GPT-4o, Claude Sonnet 4) to autonomously fix issues in GitHub repositories.
- Tool Integration: Supports custom tools and configurations for extended functionality.
- Batch Processing: Enables running multiple instances for batch processing of tasks.
- Trajectory Inspection: Provides tools to inspect and analyze agent trajectories and outputs.
- Flexible Deployment: Can be installed from source, in browser (Codespaces), or via Docker.
- Configuration Management: Supports detailed configuration files for models, tools, environments, and demonstrations.
- Tutorials & Guides: Includes comprehensive tutorials for getting started, command-line basics, and advanced usage.
- API Reference: Full documentation for run, instance, agent, and environment configurations.
Tech / Self-Hosting
- Open Source: Available under the MIT License on GitHub.
- Self-Hosted: Can be deployed locally or in cloud environments (e.g., Codespaces, Docker).
- Model Support: Compatible with various language models, including GPT-4o and Claude Sonnet 4.
- Dependencies: Requires Python and Docker for local deployment.