← Directory

Mini Swe Agent

Open Source
★ 4,882 • Updated Aug 16, 2026, 02:18 AM

Mini SWE Agent is an open-source command-line AI agent designed to autonomously solve GitHub issues or assist users directly in their terminal.

Use Cases ▼
01

Automatically solving GitHub issues

02

Interactive command-line coding assistance

03

Benchmarking AI agent performance on SWE-bench

04

Prototyping lightweight AI agent workflows

05

Educational resource for understanding agent design

What Problem It Solves ▼

Mini SWE Agent addresses the complexity and overhead typically associated with AI coding agents by providing a minimal, high-performing tool that can automatically resolve GitHub issues and assist with command-line tasks, eliminating the need for extensive configuration and infrastructure.

Key Features ▼
✓

Extremely simple (~100 lines of code)

✓

High SWE-bench Verified score (>74%)

✓

No large configuration or monorepo overhead

✓

Open source and easy to inspect

✓

Runs locally in the command line

Overview

Mini SWE Agent is an open-source command-line AI agent designed to autonomously solve GitHub issues or assist users directly in their terminal. It is built with radical simplicity in mind, consisting of approximately 100 lines of code, avoiding large configuration files or complex monorepo structures. Despite its minimal implementation, Mini SWE Agent achieves strong performance on software engineering benchmarks, scoring over 74% on SWE-bench Verified. It leverages large language models to interpret repository context, generate code changes, and execute necessary commands to resolve issues. The tool can be used to automate issue resolution, perform code modifications, and provide interactive assistance in the command line. Its lightweight nature makes it easy to inspect, modify, and integrate into existing workflows.

Pros & Cons

Pros

+

Extremely simple (~100 lines of code)

+

High SWE-bench Verified score (>74%)

+

No large configuration or monorepo overhead

+

Open source and easy to inspect

+

Runs locally in the command line

Cons

−

Limited features compared to full-featured agents

−

Requires external LLM API access

−

May need manual setup and environment configuration

−

Not a managed service; self-hosted only

−

Potential performance variability depending on model

Similar Tools

SWE-agent Aider OpenHands Devin AutoGPT GPT Engineer

More in AI Agents