Tessie MCP Server v2

MCP server rebuilt on the latest developer.tessie.com API. Summary-first tools, composite commands, and live-tested smoke scripts.

Quickstart

• Install (Smithery recommended): npx -y @smithery/cli install @keithah/tessie-mcp
• Set TESSIE_API_KEY (<https://dash.tessie.com/settings/api>) in your MCP client or .env. In Smithery UI the field appears as accessToken.
• Try in a client: “List my vehicles” → get_active_context, “Lock VIN ...” → manage_vehicle_command with confirm: true.

Tools

• get_active_context — vehicle roster with next-step guidance.
• fetch_vehicle_state — locks, climate, battery, location snapshot.
• fetch_vehicle_battery — charging-focused battery view.
• search_drives — recent drives with optional date range.
• get_driving_path — coordinate series for mapping/analysis.
• manage_vehicle_command — lock/unlock, charging, climate, speed limit, sentry, cabin overheat, seat heat/cool, flash/honk, wake.

Command safety

Destructive operations require params.confirm: true.

{
  "vin": "YOUR_VIN",
  "operation": "lock",
  "params": { "confirm": true }
}

Non-destructive actions like flash_lights / honk skip confirmation.

Local dev & tests

• Build stdio: npm run build:stdio
• Build shttp: npm run build:shttp or npm run build:all
• Tests: npm test (includes command validation)
• Smoke with live Tessie token: npm run smoke (raw client), npm run smoke:tools (MCP tools)

Smithery

• Playground/dev tunnel: npm run dev or npx @smithery/cli dev
• Transports: stdio (npm run build:stdio), shttp (npm run build:shttp, default for publish)
• Docs index: <https://smithery.ai/docs/llms.txt> ; TS quickstart: npx create-smithery@latest
• Config schema: .well-known/mcp-config (expects TESSIE_API_KEY). Server card: .well-known/mcp.json (aliases in .well-known/mcp-server.json and .well-known/mcp/server.json).
• Publish/update: npm run build:shttp → npx @smithery/cli publish (uses manifest.json). Ensure TESSIE_API_KEY is provided in user config.

Notes

• API references cached in docs/llms-full.txt and docs/tessie-api-metadata.json for offline context.
• Uses TypeScript MCP SDK and Tessie HTTPS API; all state stays in Tessie. Undo/confirmation is enforced in manage_vehicle_command.
• MCP design references: see docs/glama-links.md for glama.ai best-practice articles.
• Speed-limit operations accept speed_limit_pin (sensitive); avoid logging or sharing it.
• Optional debug logging: set TESSIE_MCP_DEBUG=1 (or true) to emit request failures with URLs/status only (no headers/API keys); retry/backoff is built-in for 429/5xx responses.
• Tessie client caches read requests (vehicles, state, battery, drives, paths, historical states) per client instance with short TTLs (15-30s), capped size (200 entries), and VIN-scoped invalidation after commands to avoid stale state while keeping token usage low.