Tesla MCP Server

Tesla MCP Server

A Model Context Protocol (MCP) server for the Tesla Fleet API. Control your Tesla and get vehicle data (location, wake up, list cars) from any MCP-capable AI assistant or agent.

Features

• list_cars — List your vehicles and get IDs for use with other tools
• get_vehicle_location — Current GPS location and Google Maps link (parking monitor style)
• wake_up — Wake a vehicle from sleep
• refresh_vehicles / debug_vehicles — Refresh list and debug info
• HTTP/SSE mode — Host as a web service; each user brings their own Tesla Developer credentials (no server-side secrets required)
• Receive and respond to texts (optional) — With Twilio configured, receive inbound SMS via webhook and use get_recent_texts / send_text so the agent can read and reply to messages

Security

• We never see or store your Tesla password. Sign-in is via Tesla’s OAuth in your browser.
• HTTP mode: Credentials and tokens are stored in memory per session only; not written to disk.
• No sensitive data in logs — We do not log tokens, full session IDs, or API response bodies.
• Before you commit: Run ./check-secrets.sh to catch accidental hardcoded secrets.

See SECURITY.md for details and how to report issues.

Quick Start (Hosted — recommended)

Use the server without running anything locally. Each user connects with their own Tesla account.

1. Add the server in your MCP client

• Server URL: https://tesla-mcp.onrender.com/sse
  (Or use your own deployed URL; see Deploy below.)

2. First time: connect your Tesla

1. Use a tool (e.g. get_setup_url) — the agent will return a link.
2. Open the link and enter your Tesla Developer Client ID and Client Secret.
3. Log in with your Tesla account when redirected.
4. On the success page, copy the connection URL (e.g. https://.../sse?token=...). Use that URL as your MCP server URL in your client so reconnects keep you logged in. Keep it private.
5. If you don’t add that URL, your client may get a new session on each reconnect and ask you to set up again.

Getting Tesla Developer credentials: Create an app at developer.tesla.com. Set the redirect URI to https://YOUR_SERVER_URL/auth/callback (e.g. https://tesla-mcp.onrender.com/auth/callback).

Render: Set Instance count to 1 (Dashboard → your service → Settings) so all requests hit the same server and your session isn’t lost.

Quick Start (Local)

Option A: HTTP server (multi-user, browser auth)

git clone https://github.com/Sara3/Tesla-MCP.git
cd Tesla-MCP
npm install
npm run build
npm run start:http

• Open http://localhost:3000 and follow the setup link to add your Tesla Developer credentials and sign in.
• In your MCP client, use Server URL: http://localhost:3000/sse (for production use HTTPS and set BASE_URL).

Option B: Stdio (single user, .env only)

For a single user with credentials in .env:

# .env
TESLA_CLIENT_ID=...
TESLA_CLIENT_SECRET=...
TESLA_REFRESH_TOKEN=...

npm run build
npm start

Configure your MCP client to run the server command (e.g. node run-mcp.js). Get a refresh token with npm run get-token.

Environment variables

SMS setup: In the Twilio console, set your phone number’s “A message comes in” webhook to https://YOUR_BASE_URL/webhooks/twilio/sms (e.g. https://tesla-mcp.onrender.com/webhooks/twilio/sms). Inbound messages are stored in memory and returned by get_recent_texts; use send_text to reply.

Never commit .env or keys/. Run ./check-secrets.sh before pushing.

Tools (MCP)