LiuMengxuan04/MiniCode

A lightweight terminal coding assistant with Claude Code-like workflow, tool loop, and TUI architecture, built for learning and experimentation. Multi-language support: TypeScript , Python and Rust implementations available now.

What it solves

MiniCode provides a lightweight, terminal-based alternative to large AI coding platforms. It solves the need for a compact, hackable coding assistant that implements a professional agentic workflow (similar to Claude Code) without the overhead of a massive codebase, making it ideal for developers who want to study, modify, or build their own custom AI tooling.

How it works

It operates on a terminal-first agent loop where the AI accepts a user request, inspects the local workspace, executes tools, and reviews file changes before writing them. The system supports a model -> tool -> model loop and can deploy up to three concurrent read-only sub-agents to assist the root agent. It integrates with the Model Context Protocol (MCP) for external tools and uses a layered memory system and context compaction techniques to manage long-session history efficiently.

Who it’s for

  • Developers seeking a lightweight terminal coding assistant for local workflows.
  • Engineers wanting a reference implementation of agentic coding architectures to study or extend.
  • Users who prefer a TUI (Terminal User Interface) with integrated tool calling and command workflows.

Highlights

  • Agentic Workflow: Supports multi-step tool execution and concurrent read-only sub-agents.
  • Context Management: Features provider-usage-first stats, auto-compacting, and context collapse to handle long sessions.
  • Extensibility: Supports local skills via SKILL.md and MCP tools/resources over stdio or HTTP.
  • TUI Experience: Full-screen terminal interface with input history, slash commands, and approval flows for file edits.
  • Multi-language Ecosystem: Available in TypeScript, Rust, Python, Go, and Java.

Related

  • Project
    XiaomiMiMo/MiMo-Code

    A terminal-native AI coding assistant with persistent memory and structured workflows for long-horizon software development.

  • Project
    anthropics/claude-code

    Claude Code is an agentic terminal-based coding tool that helps developers execute routine tasks, explain code, and manage git workflows using natural language.

  • Project
    Doorman11991/smallcode

    A terminal-native coding agent optimized for local LLMs (8B-35B parameters) that uses intelligent context management and tool routing to enable reliable coding on consumer hardware.

  • Project
    github/copilot-cli

    A terminal-based AI coding agent that allows developers to build, debug, and refactor code using natural language and GitHub integration.

  • Project
    ConardLi/easy-agent

    Easy Agent is a terminal‑based, TypeScript‑written coding assistant that streams chat with Claude, OpenAI, Gemini, Ollama and other LLMs, lets the model read/write files, run shell commands, browse the web, and be extended via plugins and multi‑agent teams. It is at release‑candidate stage, awaiting npm publication.