mini-swe-agent is a minimal AI software engineering agent from the team behind SWE-bench and SWE-agent.
Follow mini-swe-agent to get an email when its listing changes, such as new pricing or features. One digest a day at most, and you can unsubscribe anytime.
The agent class is about 100 lines of Python, it uses only bash instead of a custom tool interface, and it keeps a linear message history that is also the trajectory, so it is easy to read, modify and run with any model.
- 01
About 100 lines of Python for the core agent class, built to be hacked on
- 02
Bash as the only tool, so no tool-calling API is required and any model works
- 03
Model support via litellm, OpenRouter and Portkey, including /completion and /response endpoints
- 04
Local, Docker/Podman, Singularity/Apptainer and Bubblewrap execution environments
- 05
`mini` CLI for interactive use plus batch runs over SWE-bench and a trajectory browser
- 06
Install with `uvx mini-swe-agent`, `pipx run mini-swe-agent` or `pip install mini-swe-agent`
Open Source
Free and open-source under the MIT License; you pay only for the model API calls (via litellm, OpenRouter, Portkey or a local model) and any sandbox infrastructure you run it in.
Best for
Researchers and engineers who want a small, transparent coding agent baseline for benchmarks, fine-tuning or sandboxed batch runs, and developers who want a simple bash-only agent in the terminal.
Not ideal for
Developers who want a polished IDE experience, rich built-in tooling or a managed product with team features.
The reference minimal coding agent: tiny, model agnostic and benchmark proven, best as a hackable baseline rather than a full daily driver.
mini-swe-agent v2 is the current major version (v2.4.6 released Jul 23, 2026); the README now highlights that it powers Ramp SWE-Bench and serves as an evaluation harness for the DeepSWE benchmark, and the original SWE-agent is in maintenance mode with mini-swe-agent as its successor.
As of
Is mini-swe-agent free?+
Yes - mini-swe-agent is Open Source. Free and open-source under the MIT License; you pay only for the model API calls (via litellm, OpenRouter, Portkey or a local model) and any sandbox infrastructure you run it in.
Is mini-swe-agent open source?+
Yes - mini-swe-agent is open source. Free and open-source under the MIT License; you pay only for the model API calls (via litellm, OpenRouter, Portkey or a local model) and any sandbox infrastructure you run it in.
Who is mini-swe-agent best for?+
Researchers and engineers who want a small, transparent coding agent baseline for benchmarks, fine-tuning or sandboxed batch runs, and developers who want a simple bash-only agent in the terminal.
Who is mini-swe-agent not ideal for?+
Developers who want a polished IDE experience, rich built-in tooling or a managed product with team features.
What are the best mini-swe-agent alternatives?+
The closest mini-swe-agent alternatives on ai.dosa.dev are Jules by Google, OpenHands, SWE-agent - all listed under Autonomous & Async Agents.
Who makes mini-swe-agent?+
mini-swe-agent is developed by SWE-agent (Princeton & Stanford). It is listed in the Autonomous & Async Agents category on ai.dosa.dev.
Favorite this tool to revisit it later, or Zap it to contribute to the public vote count.