mcptoon v0.1.0
Token-efficient MCP CLI client for every AI agent.
What it does
mcptoon connects to any MCP server (stdio or HTTP) and outputs TOON (Token-Optimized Object Notation) instead of JSON — saving up to 97% tokens on tool discovery and 40-60% on tool results.
Key features
- TOON encoding — strips JSON syntax overhead, keeps real data
- Zero dependencies — pure Python stdlib, 50KB install
- Cross-platform — Windows, macOS, Linux
- Works with every agent — Claude Code, Codex, OpenCode, Cursor, CatPaw, anything that runs shell commands
- One config for all agents —
~/.mcptoon/config.json - Safety features — dangerous operation blocking, no telemetry, no credential storage
- Schema cache — 5-min TTL, avoids re-discovery
- Usage tracking — local stats, never transmitted
Token savings
| Operation | JSON | mcptoon TOON | Savings |
|---|---|---|---|
| Tool discovery (96 tools) | ~2,000 | ~60 | 97% |
| Tool result (structured) | ~800 | ~350 | 56% |
| Tool result (raw HTML) | ~1,000 | ~900 | 10% |
Install
pip install mcptoonQuick start
mcptoon init
mcptoon add fetch --stdio npx -y @modelcontextprotocol/server-fetch
mcptoon manifest --toon
mcptoon call fetch fetch '{"url":"https://example.com"}' --toonFull Changelog: https://github.com/activeing123/mcptoon/commits/v0.1.0