SWE-agent SWE-agent
SWE-agent is a research-grade framework that enables language models to autonomously fix issues in GitHub repositories by utilizing a specialized Agent-Computer Interface. It is designed to navigate codebases, execute commands, and run tests to resolve software engineering tasks without human intervention.
The project is currently in maintenance-only mode, as the development team has shifted focus to mini-swe-agent, a simpler and more flexible successor. Recent technical milestones include integration with SWE-ReX for parallel execution and compatibility updates for SWE-bench benchmarks.
As of
- Agent-Computer Interface (ACI) for structured code navigation and editing
- Docker-based sandboxing for secure, isolated code execution
- Configurable YAML-based pipelines for custom tool definitions and agent behavior
- Support for multimodal models to process images from GitHub issues
- EnIGMA mode for autonomous vulnerability discovery and cybersecurity challenges
ML researchers studying AI agent behavior, engineering teams performing batch autonomous issue resolution on self-hosted infrastructure, and security researchers conducting CTF challenges.
Developers seeking an interactive, day-to-day coding companion or non-technical users requiring an out-of-the-box GUI or managed service.
The software is free and open-source under the MIT license; users pay for the underlying LLM API usage, which typically costs $0.50 to $3.00 per issue depending on the model and task complexity.
SWE-agent is a foundational project that defined modern autonomous coding agent research. While it remains a powerful, high-performance tool for researchers and developers comfortable with its technical setup, users seeking new projects are strongly encouraged by the creators to use the more efficient mini-swe-agent.
Is SWE-agent free?
Yes - SWE-agent is Open Source. The software is free and open-source under the MIT license; users pay for the underlying LLM API usage, which typically costs $0.50 to $3.00 per issue depending on the model and task complexity.
What is SWE-agent best for?
ML researchers studying AI agent behavior, engineering teams performing batch autonomous issue resolution on self-hosted infrastructure, and security researchers conducting CTF challenges.
Who makes SWE-agent?
SWE-agent is developed by Princeton NLP and Stanford University. It is listed in the Autonomous & Async Agents category on ai.dosa.dev.
Sign in with GitHub or Google to Zap tools, vote, and get a README badge showing your stack.
Content on this page is AI-generated. Please verify details with the vendor's website for accuracy.