SWE-agent/mini-swe-agent

What is it?

Key features

• 01Agent core is about 100 lines of Python with no heavy dependencies.
• 02Uses bash as the only tool and does not require tool-calling support from the model.
• 03Runs each action with subprocess.run, so execution can be sandboxed via Docker, Podman, Apptainer or Bubblewrap.
• 04Works with models through LiteLLM, OpenRouter and Portkey.
• 05Includes a CLI, Python bindings, SWE-bench batch inference and a trajectory browser.

Use cases

• Running a quick AI coding assistant from the terminal against a local project.
• Using a minimal, hackable agent as a baseline for agent or model research.
• Evaluating language models on SWE-bench with a bash-only scaffold.
• Running agents inside sandboxed containers for safer automated code changes.

Quick start

1. 1

   Install and try the CLI in an isolated environment with uv

   pip install uv && uvx mini-swe-agent
2. 2

   Or install the CLI and Python bindings in your current environment

   pip install mini-swe-agent
3. 3

   Run the CLI

   mini

Things to know

• !Model access requires an API key or a compatible model endpoint, which the README does not detail beyond litellm and openrouter support.
• !Full SWE-bench-style runs and sandboxed environments need extra setup such as Docker, Podman, or Apptainer.

Topics

FAQ

What is mini-swe-agent?

A minimal AI software engineering agent that solves GitHub issues or helps in the command line, using only bash and about 100 lines of core Python. mini-swe-agent is a small AI agent for software engineering tasks. It can work on GitHub issues or act as a command-line assistant. Its agent class is roughly 100 lines of Python, and it depends on no heavy frameworks.

Is mini-swe-agent free to use? What license does it have?

mini-swe-agent is open source under the MIT license. Check the license terms before using it in commercial products.

Is mini-swe-agent still maintained?

Actively maintained. The last commit to mini-swe-agent was on Oct 19, 2026.

What can I use mini-swe-agent for?

Running a quick AI coding assistant from the terminal against a local project. Using a minimal, hackable agent as a baseline for agent or model research. Evaluating language models on SWE-bench with a bash-only scaffold. Running agents inside sandboxed containers for safer automated code changes.

How do I install mini-swe-agent?

Install and try the CLI in an isolated environment with uv: pip install uv && uvx mini-swe-agent Or install the CLI and Python bindings in your current environment: pip install mini-swe-agent Run the CLI: mini

Where can I get mini-swe-agent?

The source code and installation instructions are on GitHub: https://github.com/SWE-agent/mini-swe-agent