SWE-agent/mini-swe-agent

Health v2 · maintenance only

Adoption not part of the score

6779 stars · 946 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

A radically minimal AI software engineering agent (~100 lines of core Python) that solves GitHub issues and assists in the command line, scoring over 74% on SWE-bench Verified. It supports all LLMs via LiteLLM and runs in local, Docker, Podman, Singularity, and other sandboxed environments.

Use cases

• automatically fix github issues with an ai agent
• run a coding agent from the command line
• evaluate llms on swe-bench
• build a minimal ai agent from scratch
• let an ai edit and run code in a sandbox
• benchmark coding agents cheaply

When to choose

• you want a simple, hackable coding agent without heavy configs
• you need a fast SWE-bench evaluation harness
• you want to learn how coding agents work internally
• you need broad model support via LiteLLM/OpenRouter

When to avoid

• you need rich IDE integration or interactive UX like Claude Code
• you want extensive built-in tooling, plugins, and configs
• you need non-Python or non-shell agent environments

Facets

cli-tool · maturity active

agent-framework llm-inference cli developer-tools large-language-models developer-tools python cli cross-platform coding-agent swe-bench litellm github-issues minimal-agent software-engineering ai-agents command-line docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="SWE-agent/mini-swe-agent")