LettuceMeet

Unofficial CLI + MCP tools for LettuceMeet.

Create polls, read responses, find meeting overlaps -- all from the terminal. Designed for AI agents that want to automate scheduling without a browser.

Why

LettuceMeet has no public API. The official platform only works through a web browser. This project reverse-engineers the internal GraphQL API so that AI agents can schedule meetings directly from the command line.

Built by someone who believes every web app should be automatable from a terminal. CLI-first, agent-friendly, zero bloat.

How it works

A Python CLI (argparse + requests) talks to api.lettucemeet.com/graphql using JWT auth. An MCP server wrapper exposes the same operations as callable tools for any MCP-compatible agent (Pi, Claude Code, Codex, etc.).

User / Agent -> CLI / MCP -> GraphQL -> lettucemeet.com

Quick setup

uv sync
uv run python main.py login "<token>"   # one-time: paste token from user

Token -- user runs this on lettucemeet.com (logged in) > DevTools Console:

copy(localStorage.getItem('akoko:session_token'))

Commands

Token resolution: --token flag > LETTUCEMEET_TOKEN env var > data/session.json

MCP server

{
  "mcpServers": {
    "lettucemeet": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to", "python", "src/lettucemeet_cli/mcp_server.py"]
    }
  }
}

Tools: create_poll, show_event, respond_to_poll, compute_overlap, delete_event

Agent skill

bash scripts/install-skill.sh --link

Agents auto-discover and can run CLI commands autonomously.

Project

src/lettucemeet_cli/   # Python package (6 modules, 42 tests)
scripts/               # Skill installer
docs/                  # Architecture, MCP docs

uv run pytest          # 42 tests

License

MIT