# mini-SWE-agent review

> mini-SWE-agent is a minimal Python AI agent, about 100 lines for its agent class, that solves GitHub issues or helps in your command line.

Devtoolwatch Board, the week of Oct 25–31, 2026. Canonical page: https://index.devtoolwatch.com/tools/mini-swe-agent/

## TL;DR

mini-SWE-agent is a minimal AI software engineering agent from the team behind SWE-bench and SWE-agent. It is for developers who want a simple command line agent and for researchers running benchmarks, fine-tuning or RL. Its only tool is bash, it keeps a linear history, and it runs each action independently.

## mini-SWE-agent at a glance

- **Category:** [Coding agents](https://index.devtoolwatch.com/categories/coding-agents/)
- **Maintained by:** SWE-agent (organization, [@SWE-agent](https://github.com/SWE-agent) on GitHub)
- **License:** MIT
- **Language:** Python
- **Repository:** [github.com/SWE-agent/mini-swe-agent](https://github.com/SWE-agent/mini-swe-agent)
- **Website:** [mini-swe-agent.com](https://mini-swe-agent.com)
- **PyPI package:** [mini-swe-agent](https://pypi.org/project/mini-swe-agent/) 2.4.6
- **Latest release:** [v2.4.6](https://github.com/SWE-agent/mini-swe-agent/releases/tag/v2.4.6) on GitHub, Jul 23, 2026
- **First published:** Jun 28, 2025

## What mini-SWE-agent does

- mini-SWE-agent solves GitHub issues or helps you with tasks in your command line.
- mini-SWE-agent uses bash as its only tool and does not need the tool-calling interface of language models.
- mini-SWE-agent runs in local environments, docker/podman, singularity/apptainer, bubblewrap, contree and more.
- mini-SWE-agent supports models via litellm, openrouter and portkey, including /completion and /response endpoints.
- mini-SWE-agent includes a trajectory browser, batch inference and Python bindings.
- mini-SWE-agent keeps a linear message history, so the trajectory equals what the model is prompted with.

### How it works

mini-SWE-agent runs as the mini CLI or through Python bindings, such as a DefaultAgent with a LitellmModel and LocalEnvironment. Each step appends to a linear message history. The model issues bash commands, which run with subprocess.run as independent actions instead of a stateful shell session. That design lets you swap in docker exec to run actions in sandboxes.

## How to install mini-SWE-agent

**Try the CLI with uv**

```
pip install uv && uvx mini-swe-agent
```

**With pip**

```
pip install mini-swe-agent
```

### Requirements

- Python 3.10 or newer
- Access to a language model through litellm, openrouter, portkey or a similar provider

### First steps

1. Install the package with pip install mini-swe-agent.
2. Run the CLI with the mini command.
3. Read the quick start guide and the global configuration docs to set up models and settings.

## Who mini-SWE-agent is for

- Developers who want a quick command line agent that works locally
- Researchers running sandboxed benchmark evaluations such as SWE-bench
- Teams doing fine-tuning or RL who want to avoid overfitting to one agent scaffold
- Developers who want a hackable agent with very simple control flow

### Limitations

- mini-SWE-agent has no tools other than bash, so specific tasks rely on the model using the shell.
- mini-SWE-agent does not keep a stateful shell session; every action runs independently.
- For experimenting with custom tool sets or history processors, the README recommends SWE-agent instead.
- mini-SWE-agent v2 changed from v1; v1 users should read the migration guide.

## Usage data (the week of Oct 25–31, 2026)

- Board rank: #236 overall, #79 in Coding agents, momentum score 39
- Downloads that week: 3.6M (−3% vs 4-week average)
- New GitHub stars per week (last 12 weeks): Aug 9: 212; Aug 16: 165; Aug 23: 163; Aug 30: 131; Sep 6: 492; Sep 13: 331; Sep 20: 207; Sep 27: 189; Oct 4: 194; Oct 11: 236; Oct 18: 210; Oct 25: 196

## mini-SWE-agent alternatives

| Rank | Tool | Stars | Downloads/wk |
|---|---|---|---|
| #1 | [Cindy](https://index.devtoolwatch.com/tools/makecindy-cindy/) | 8.8k | – |
| #2 | [OpenMausBot](https://index.devtoolwatch.com/tools/milind-soni-openmausbot/) | 9.6k | – |
| #3 | [Orca](https://index.devtoolwatch.com/tools/stablyai-orca/) | 111.7k | – |
| #4 | [OpenBot](https://index.devtoolwatch.com/tools/copilotkit-openbot/) | 9.1k | – |
| #5 | [Agent Reach](https://index.devtoolwatch.com/tools/panniantong-agent-reach/) | 107.2k | – |
| #6 | [Skill Recorder](https://index.devtoolwatch.com/tools/microsoft-skill-recorder/) | 5.6k | – |

## mini-SWE-agent FAQ

### Is mini-SWE-agent free and open source?

Yes. mini-SWE-agent's source code is on GitHub (SWE-agent/mini-swe-agent) under the MIT license, so you can use it for free.

### How do I install mini-SWE-agent?

Run `pip install uv && uvx mini-swe-agent`. The mini-swe-agent package needs Python 3.10 or newer. The install section above has the other options from the project's README.

### Is mini-SWE-agent still maintained?

Yes. The SWE-agent/mini-swe-agent repository was last updated on Oct 19, 2026, and the latest release, v2.4.6 on GitHub, came out on Jul 23, 2026.

### How popular is mini-SWE-agent?

mini-SWE-agent has 9k GitHub stars as of Nov 1, 2026. It gained 196 new stars in the week of Oct 25–31, 2026, against a 12-week average of 227 a week. Its main package had 3.6M downloads that week, −3% against its own 4-week average. That made it #79 among AI coding agents on the Devtoolwatch Board.

### mini-SWE-agent vs Claude Code: how do they compare?

mini-SWE-agent and Claude Code are both tracked by Devtoolwatch as AI coding agents. mini-SWE-agent has 9k GitHub stars and Claude Code has 154.1k. In the week of Oct 25–31, 2026, mini-SWE-agent's main package had 3.6M downloads and Claude Code's had 11.6M. Claude Code is Anthropic's agentic coding tool that reads your codebase, edits files, runs commands, and handles git workflows through natural language. [Claude Code review](https://index.devtoolwatch.com/tools/claude-code/)

### What are the best alternatives to mini-SWE-agent?

Other AI coding agents on the Devtoolwatch Board in the week of Oct 25–31, 2026, by momentum: Cindy (#1), OpenMausBot (#2), Orca (#3), OpenBot (#4), Agent Reach (#5). The alternatives table above has their numbers.

### How does mini-SWE-agent compare to Claude Code?

mini-SWE-agent's README says it starts much faster than Claude Code. It also links a DeepSWE evaluation reporting that mini-SWE-agent beats Claude Code and Codex on that benchmark.

### Which models does mini-SWE-agent work with?

mini-SWE-agent supports all models via litellm, openrouter, portkey and more. Because it uses only bash and not the tool-calling interface, the README says it can run with any model.

### Should I use SWE-agent or mini-SWE-agent?

The README recommends mini-SWE-agent as the default choice. It suggests SWE-agent only if you want to experiment with different tool sets or history processors.

### Can mini-SWE-agent run in sandboxes?

Yes. mini-SWE-agent supports docker/podman, singularity/apptainer, bubblewrap and contree. Actions run through subprocess.run, which can be swapped for docker exec.

### Can I use mini-SWE-agent from Python code?

Yes. mini-SWE-agent has Python bindings. You build a DefaultAgent with a model such as LitellmModel and an environment such as LocalEnvironment, then call agent.run with a task.

## Sources

1. [SWE-agent/mini-swe-agent on GitHub](https://github.com/SWE-agent/mini-swe-agent)
2. [mini-SWE-agent README](https://github.com/SWE-agent/mini-swe-agent/blob/main/README.md)
3. [mini-swe-agent on PyPI](https://pypi.org/project/mini-swe-agent/)
4. [mini-SWE-agent website](https://mini-swe-agent.com)
5. [Release v2.4.6 (Jul 23, 2026)](https://github.com/SWE-agent/mini-swe-agent/releases/tag/v2.4.6)
6. [Release v2.4.5 (Jul 6, 2026)](https://github.com/SWE-agent/mini-swe-agent/releases/tag/v2.4.5)
7. [Release v2.4.4 (Jul 2, 2026)](https://github.com/SWE-agent/mini-swe-agent/releases/tag/v2.4.4)
8. [Hacker News: Show HN: Mini-swe-agent achieves 65% on SWE-bench in 100 lines of python](https://news.ycombinator.com/item?id=44682897)
9. [Devtoolwatch Board, the week of Oct 25–31, 2026](https://index.devtoolwatch.com/)