SWE-agent

Active

Description

SWE-agent takes a GitHub issue and automatically generates fixes using your LLM of choice, also applicable to cybersecurity auditing and competitive coding. NeurIPS 2024 paper.

Key Features

State-of-the-art performance on SWE-bench among open-source software engineering agents
Supports multiple LLM backends including GPT-4o and Claude Sonnet 4
Configurable via a single YAML file with full documentation
Includes EnIGMA mode for offensive cybersecurity (CTF) challenge solving
Free-flowing agent design that maximizes LM autonomy for tool use
Simple and hackable architecture ideal for research and experimentation

Use Cases

💡 Automatically fix issues in real GitHub repositories

💡 Find cybersecurity vulnerabilities via the EnIGMA CTF mode

💡 Solve competitive programming challenges

💡 Research and experiment with LLM-based software engineering agents

💡 Batch benchmark evaluation on SWE-bench datasets

Quick Start

1. Install: git clone https://github.com/SWE-agent/SWE-agent && cd SWE-agent && pip install -e .
2. Set API key: export OPENAI_API_KEY=your_key
3. Run: python -m sweagent run --problem_file <issue_file> --agent_config default.yaml
4. Or try in browser via GitHub Codespaces

Visit GitHub Visit Website

Related Projects

AutoCodeRover

3.1k · Python

Stale

AutoCodeRover is a project structure-aware autonomous software engineer agent that achieves automated program repair and issue resolution by understanding the overall codebase architecture.

codingpythonagent +2

DeepCode

15.8k · Python

Normal

DeepCode is an open agentic coding platform supporting Paper2Code, Text2Web, and Text2Backend, leveraging agent technology for automated software development workflows.

codingpythonllm +2

Mini SWE-Agent

5.3k · Python

Active

Mini SWE-Agent is a minimalist AI agent in just 100 lines of code that solves GitHub issues or assists developers in the command line, demonstrating core coding agent capabilities with minimal implementation.

codingpythonagent +2

SWE-bench

5.2k · Python

Normal

SWE-bench is a benchmark for evaluating language models on real-world GitHub issue resolution, featuring genuine problems from popular Python repositories, now a core standard for measuring AI coding agent capabilities.

evaluationpythoncoding +2

AI 编程Coding AgentCLI

AI Coding Agents Deep Dive: Architecture Trade-offs from CLI to IDE-Integrated

A deep architectural comparison of seven open-source coding agents across three paradigms — CLI-first, IDE-integrated, and fully autonomous — examining context management, tool access, and autonomy levels to help you pick the right tool for each development scenario.