mcptoon
One zero-dependency CLI for every MCP server and agent skill. Token optimization, tool discovery and context compression: 71,929 -> 581 tokens (-99.2%, measured
What is mcptoon?
One zero-dependency CLI for every MCP server and agent skill. Token optimization, tool discovery and context compression: 71,929 -> 581 tokens (-99.2%, measured), schemas stay out of context. One config for Claude Code, Codex, Cursor, every agent. 341KB, pure Python. | 零依赖 CLI:管所有 MCP 工具与技能,工具发现省 99.2% token。
Overview of mcptoon
What it is
mcptoon is a zero-dependency CLI tool for managing MCP servers and agent skills. It provides token optimization, tool discovery, and context compression. The tool reduces tool schema context from 71,929 tokens to 581 tokens, a 99.2% reduction, as measured. It is 534KB, written in pure Python, and licensed under Apache-2.0.
Key features
- Token optimization: compresses tool schemas by up to 99.2% (measured on 255 tools) and skills by up to 99.99%.
- Tool discovery: finds MCP servers already installed and searches for new ones by plain words, then installs and calls them.
- Context compression: reduces document context from 40k to 17k tokens per question.
- Unified governance: works with 103+ desktop and CLI agents via two legs: CLI (agents that run a shell) and serve (agents that mount MCP).
- Automatic agent configuration:
mcptoon quickstartdetects MCP servers and auto-configures agents, including self-installing the/mcptoonskill. - Built-in skill: agents can load
/mcptoonto drive all features via natural language.
Use cases
- Adding many MCP tools and skills without flooding the agent's context window.
- Managing multiple agents (Claude Code, Codex, Cursor, Windsurf, etc.) with a single configuration.
- Discovering and calling the right MCP tool from a large collection using plain words.
- Reducing token usage and costs in AI agent interactions.
Getting started
bashpip install mcptoonmcptoon quickstartFor alternative installation methods (pipx, uvx, source, Nix), see the repository README.
License and pricing
Licensed under Apache-2.0. Open source.
Based on the project's public README and repository metadata.