Repolode
LiuMengxuan04

LiuMengxuan04/MiniCode

A lightweight terminal coding assistant with a Claude Code-like tool loop and TUI, built as a small codebase for learning and experimentation.

Open-source alternative to:Claude Code

Stars
1,161
Forks
125
Open issues
16
Last commit
29 days ago

What is it?

MiniCode is a terminal-first coding assistant for local development. It takes a user request, inspects the workspace, calls tools such as file reading, search, editing and command execution, reviews file changes before writing, and returns a final response in the same session.

The project aims to reproduce the workflow and architectural ideas of Claude Code in a much smaller implementation. It includes a full-screen TUI with transcript scrolling, slash commands, approval flows, per-project session persistence, and context management such as auto-compact. It also supports local skills defined by SKILL.md and MCP tools over stdio or remote HTTP.

Besides the TypeScript repository, the same project has Rust, Python, Go and Java implementations maintained in separate repositories. It is primarily suited to people who want to study how a coding agent is built or to customize their own tooling.

Key features

  • 01Multi-step tool calls in a single turn, forming a model-tool-model loop
  • 02Full-screen terminal UI with input history, slash command menu and approval flows
  • 03Review-before-write file edits with path and command permission checks
  • 04Per-project session persistence with resume, rename, fork and compact
  • 05Local skills via SKILL.md and MCP tools, resources and prompts over stdio or HTTP
  • 06Up to 3 concurrent read-only sub-agents coordinated by a root agent

Use cases

  • Studying how a Claude Code-style coding agent loop and TUI are structured
  • Building a custom terminal coding assistant on a small, readable codebase
  • Experimenting with MCP servers and local SKILL.md skills in a terminal agent
  • Running an offline demo of the agent workflow with a mock model mode

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/LiuMengxuan04/MiniCode.git
  2. 2

    Enter the project directory

    cd MiniCode
  3. 3

    Install dependencies

    npm install
  4. 4

    Run the installer and enter model name, ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN, then start with minicode

    npm run install-local

Things to know

  • Requires an Anthropic-compatible API endpoint and auth token to use a real model.
  • Designed as a small learning and experimentation codebase rather than a full production platform.

In collections

Topics

FAQ

What is MiniCode?
A lightweight terminal coding assistant with a Claude Code-like tool loop and TUI, built as a small codebase for learning and experimentation. MiniCode is a terminal-first coding assistant for local development. It takes a user request, inspects the workspace, calls tools such as file reading, search, editing and command execution, reviews file changes before writing, and returns a final response in the same session.
Is MiniCode free to use? What license does it have?
MiniCode is open source under the MIT license. Check the license terms before using it in commercial products.
Is MiniCode still maintained?
Actively maintained. The last commit to MiniCode was on Oct 9, 2026.
What can I use MiniCode for?
Studying how a Claude Code-style coding agent loop and TUI are structured Building a custom terminal coding assistant on a small, readable codebase Experimenting with MCP servers and local SKILL.md skills in a terminal agent Running an offline demo of the agent workflow with a mock model mode
How do I install MiniCode?
Clone the repository: git clone https://github.com/LiuMengxuan04/MiniCode.git Enter the project directory: cd MiniCode Install dependencies: npm install Run the installer and enter model name, ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN, then start with minicode: npm run install-local
Where can I get MiniCode?
The source code and installation instructions are on GitHub: https://github.com/LiuMengxuan04/MiniCode

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects