mini-swe-agent
View on GitHubThe 100 line AI agent that solves GitHub issues or helps you in your command line. Radically simple, no huge configs, no giant monorepo—but scores >74% on SWE-bench verified!
mini-swe-agent is a ~100-line Python software engineering agent that solves GitHub issues and runs as a CLI coding assistant. It uses only bash, has no tool-calling scaffolding, supports any LiteLLM model, and scores >74% on SWE-bench verified.
Use Cases
solve GitHub issues automaticallycommand line coding assistantSWE-bench verified evaluationbatch inference over software taskssandboxed code execution via docker/podmanbaseline scaffold for RL/FT agent trainingtrajectory inspection and debuggingPython bindings for embedding an agent
Built With
- Language
- Python
- Frameworks
- litellm · openai · textual · pydantic · typer · jinja2 · docker · podman · singularity · modal · swe-rex · pytest · rich · prompt_toolkit
Tags
coding-agent · swe-bench · cli · minimal · bash · agent · llm · sandboxing · python · open-source · benchmark · automation