A lightweight, terminal-based coding assistant inspired by Claude Code's workflow and architecture, MiniCode targets developers seeking a simple, understandable, and extensible system for learning, experimentation, or building custom tooling within their local development environment. The primary benefit is a compact, efficient agent loop accessible directly from the terminal, designed for speed and simplicity.

How It Works

The core operates on a terminal-first agent loop: processing user requests, inspecting the workspace, invoking tools, reviewing file changes, and returning responses within a single session. Its intentionally compact codebase, featuring a TUI and a model-tool-model loop, prioritizes clarity and ease of modification over feature breadth. This design facilitates understanding of agent control flow and tool integration.

Quick Start & Requirements

• Primary install/run command: npm install followed by npm run install-local.
• Non-default prerequisites: Node.js (via npm), API keys for model providers (e.g., ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN). Configuration is stored in ~/.mini-code/settings.json and ~/.mini-code/mcp.json. Ensure ~/.local/bin is in your system's PATH.
• Links: Local docs/index.html or GitHub Pages: https://liumengxuan04.github.io/MiniCode/.

Highlighted Details

• Multi-language implementations: Reference TypeScript, alongside Rust (MiniCode-rs) and Python branches, offering different context management and features.
• Extensive built-in tools: Includes file operations (list_files, read_file, write_file, edit_file), command execution (run_command), web interaction (web_fetch, web_search), and agent control (ask_user, load_skill).
• Advanced TUI interactions: Features input history, transcript scrolling, slash commands, and sophisticated approval prompts with selection, shortcuts, and "reject with guidance" options.
• Model Context Protocol (MCP) support: Enables integration with external tool providers via stdio, with automatic tool registration and flexible framing negotiation (Content-Length or newline-delimited JSON).

Maintenance & Community

The project provides links to contribution guidelines, a roadmap, and architecture documentation, emphasizing its design for learning and extensibility. No specific community channels (e.g., Discord, Slack) or maintainer details are listed.

Licensing & Compatibility

The license type is not explicitly stated in the provided README text.

Limitations & Caveats

The Python implementation is still under development with features being shaped. Remote MCP authentication currently relies on token-based methods for simplicity, deferring full OAuth automation. The project prioritizes a small, hackable architecture, suggesting it may not encompass the full feature set of larger, more established platforms.