Cassini Mission Plan MCP

A read-only Model Context Protocol server over the Cassini-Huygens mission dataset (~62k activity rows). Built as an AI workshop demo: clean, readable TypeScript that runs live in Claude Desktop.

Live endpoint: https://cassini-mission-plan.redfour.workers.dev

What it does

Exposes 7 MCP tools so an LLM client can query the Cassini master_plan table:

Related MCP server: hiveecho

Architecture

• Runtime: Cloudflare Workers (TypeScript)

• Data: Cloudflare D1 (SQLite) — master_plan table + FTS5 virtual table

• Transport: hand-rolled MCP-over-HTTP (JSON-RPC 2.0 over plain HTTP POST)

• Validation: zod schemas per tool; errors surface as JSON-RPC error objects

See docs/ARCHITECTURE.md for the full design.

Project setup

npm install

Run tests

npm test                  # Jest — all specs against in-memory SQLite

The deploy spec (spec/deploy-and-initialize.spec.ts) needs a live URL:

DEPLOY_URL=https://cassini-mission-plan.redfour.workers.dev npm test

Deploy (one-time setup)

1. Create the D1 database:

   npx wrangler d1 create cassini

   Copy the database_id into wrangler.toml.

2. Import the data (generates data/cassini.d1.sql from data/cassini.db):

   node scripts/import.js      # or: npx ts-node scripts/import.ts
   npx wrangler d1 execute cassini --remote --file=data/cassini.d1.sql

3. Deploy:

   npm run deploy

Local dev

npm run dev   # wrangler dev (uses local D1)

Docs