mini-SWE-agent

What it does

mini-SWE-agent is a minimal AI coding agent that takes a natural-language task, such as implementing a Sudoku solver or fixing failing unit tests, and lets a language model complete it by issuing bash commands in a local or sandboxed environment.

It can be used through the mini command line interface, the mini-extra utilities, or Python bindings such as DefaultAgent with LitellmModel and LocalEnvironment. It has no tools other than bash, does not use the model's tool-calling interface, and runs each action with subprocess.run, with a completely linear message history that doubles as the trajectory.

It suits developers and researchers who want a hackable baseline agent rather than a UI-heavy product. You need Python and access to a language model through a supported provider, which you configure the first time you run mini. Its documentation points to swe-agent for custom tool sets or history processors.

What it can do

• Runs a bash-only agent loop that lets the language model execute shell commands to complete tasks.
• Executes each action independently with subprocess.run, supporting local and sandboxed environments such as docker or podman.
• Keeps a completely linear message history that doubles as the trajectory for debugging and fine-tuning.
• Provides a mini command-line interface and mini-extra utilities for running tasks.
• Offers Python bindings through DefaultAgent with LitellmModel and LocalEnvironment.
• Supports models via litellm, openrouter, and portkey, including /completion and /response endpoints.

Before you choose

• A language model is required, and costs for the chosen provider are separate from the package.
• Designed as a minimal baseline; its documentation recommends swe-agent for custom tool sets or history processors.
• Results depend on the language model used, and the recommended model names in the docs may change.

Listed under

• Coding
• Coding agents
• AI agent
• Benchmark
• Benchmarks
• Coding agent
• Command line
• Developer tools

Get started

Install the mini-swe-agent Python package, then run the mini command; model setup runs on first use.

1. Install the mini-swe-agent package with pip, uv, or pipx.
2. Run the mini command; if setup was skipped, run mini-extra config setup.
3. Include the provider in the model name, for example anthropic/claude-... or openai/gpt-5.4.

Model provider charges may apply separately; check your chosen provider's pricing.

We couldn’t find published prices for mini-SWE-agent.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Overview - mini-SWE-agent documentation ↗mini-swe-agent.com
• Quick start - mini-SWE-agent documentation ↗mini-swe-agent.com

Company website

mini-swe-agent.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.