OSS Alternative - Discover Top Open Source Alternatives to Popular Software

SWE-agent/mini-swe-agent

A radically simple yet highly performant AI agent for software engineering, capable of solving GitHub issues and assisting in command-line tasks.

Core Features

Minimalist design with ~100 lines of Python for the core agent.

Achieves over 74% on the SWE-bench verified benchmark, outperforming more complex agents.

Supports flexible deployment across local, containerized (Docker), and sandboxed environments.

Compatible with various LLMs via LiteLLM, OpenRouter, and other endpoints.

Utilizes only Bash as its tool, ensuring broad LLM compatibility and simplified sandboxing.

Detailed Introduction

mini-swe-agent is a groundbreaking, minimalist AI agent developed by the creators of SWE-bench and SWE-agent. It challenges the notion that powerful AI agents require complex architectures, demonstrating that a core agent with just ~100 lines of Python can achieve over 74% performance on the SWE-bench verified benchmark. Designed for software engineering tasks, it excels at resolving GitHub issues and providing command-line assistance. Its simplicity, broad compatibility with LLMs and deployment environments, and high performance have led to its wide adoption across leading tech companies and academic institutions, making it an ideal baseline for research and a hackable tool for developers.