# mini-SWE-agent [Coding agents](https://www.agentlist.io/categories/cli-agent)SWE-agent[Last checked Nov 12, 2026](https://www.agentlist.io/changelog) mini-SWE-agent is a roughly 100-line open-source coding agent from the SWE-bench team that fixes GitHub issues or works in your terminal with any model. [Visit mini-SWE-agent](https://mini-swe-agent.com/) - [mini-swe-agent.com](https://mini-swe-agent.com/) - [github.com/SWE-agent/mini-swe-agent](https://github.com/SWE-agent/mini-swe-agent) - Open source (MIT) - Terminal - Any model - SWE-bench ## At a glance **Who it's for**: Researchers and engineers who want a minimal, hackable coding agent and a benchmark-grade baseline. **What does it do?**: A minimal AI software engineering agent that solves GitHub issues or helps on the command line. The agent class is about 100 lines of Python and all it needs is bash. The README reports over 74% on SWE-bench verified. [github.com (source for What does it do?)](https://github.com/SWE-agent/mini-swe-agent) [mini-swe-agent.com (source for What does it do?)](https://mini-swe-agent.com/) **Where does it run?**: Locally on your machine, or in docker/podman, singularity/apptainer, bubblewrap and other sandboxed environments. [github.com (source for Where does it run?)](https://github.com/SWE-agent/mini-swe-agent) **What does it cost?**: Free and open source under MIT. You pay your own model provider. [github.com (source for What does it cost?)](https://github.com/SWE-agent/mini-swe-agent) **Which models does it work with?**: Any model. It uses LiteLLM for model access and needs no tool-calling support. [github.com (source for Which models does it work with?)](https://github.com/SWE-agent/mini-swe-agent) **Who makes it?**: The SWE-agent and SWE-bench team (GitHub org SWE-agent), researchers from Princeton and Stanford. It supersedes their earlier SWE-agent. [github.com (source for Who makes it?)](https://github.com/SWE-agent/mini-swe-agent) [github.com (source for Who makes it?)](https://github.com/SWE-agent/SWE-agent) ## Common questions What is mini-SWE-agent? mini-SWE-agent is a roughly 100-line open-source coding agent from the SWE-bench team that fixes GitHub issues or works in your terminal with any model. Who is mini-SWE-agent for? Researchers and engineers who want a minimal, hackable coding agent and a benchmark-grade baseline. How much does mini-SWE-agent cost? Free and open source under MIT. You pay your own model provider. Checked against the vendor's pages on Nov 12, 2026. Plans change often, so confirm before you decide. Does mini-SWE-agent run locally or in the cloud? mini-SWE-agent runs in the vendor's cloud. You use it through chat apps and an API. What are alternatives to mini-SWE-agent? See the Coding agents shortlist for the other options. [See all CLI agents](https://www.agentlist.io/categories/cli-agent) --- Source: [mini-SWE-agent: pricing & alternatives | agentlist.io](https://www.agentlist.io/systems/mini-swe-agent). This is the public page rendered as Markdown; interactive controls require the website.