iceener

Tesla MCP Server

Community iceener
Updated

MCP Server for interacting with Tesla (via Tessie API). Written in TypeScript, Node and Hono.dev

Tesla MCP Server

Streamable HTTP MCP server for Tesla vehicle control via the Tessie API.

Author: overment

[!WARNING]You connect this server to your MCP client at your own responsibility. Language models can make mistakes, misinterpret instructions, or perform unintended actions. Always verify commands before execution, especially for actions like unlocking, opening trunks, or sending navigation destinations.

The HTTP layer is designed for convenience during development, not production-grade security. If deploying remotely, harden it: proper token validation, secure storage, TLS termination, strict CORS/origin checks, rate limiting, and audit logging.

Notice

This repo works in two ways:

  • As a Node/Hono server for local workflows
  • As a Cloudflare Worker for remote interactions

Features

  • ✅ State — Battery, range, location, climate, doors, charging status
  • ✅ Commands — Lock/unlock, climate, trunks, sentry, navigation
  • ✅ Location-aware — GPS coordinates for context-aware interactions
  • ✅ Dual Runtime — Node.js/Bun or Cloudflare Workers

Design Principles

  • LLM-friendly: Two unified tools, not 1:1 API mirrors
  • Watch-ready: Designed for AI agents with location context
  • Secure: Tessie API key stored as secret, clients use separate bearer token
  • Clear feedback: Detailed command results and vehicle state

Installation

Prerequisites: Bun, Tessie Account.

Ways to Run (Pick One)

  1. Local Development — Standard setup with bearer token auth
  2. Cloudflare Worker (wrangler dev) — Local Worker testing
  3. Cloudflare Worker (deploy) — Remote production

1. Local Development — Quick Start

  1. Get Tessie credentials:

    • Visit developer.tessie.com
    • Go to Developer Settings → Generate Access Token
    • Copy your access token
    • Note your vehicle's VIN
  2. Configure environment:

cd tesla-mcp
bun install
cp .env.example .env

Edit .env:

PORT=3000
AUTH_ENABLED=true
AUTH_STRATEGY=bearer

# Generate with: openssl rand -hex 32
BEARER_TOKEN=your-random-auth-token

# Tessie credentials
TESSIE_ACCESS_TOKEN=your-tessie-access-token
TESSIE_VIN=your-vehicle-vin
  1. Run:
bun dev
# MCP: http://127.0.0.1:3000/mcp

Claude Desktop / Cursor:

{
  "mcpServers": {
    "tesla": {
      "command": "npx",
      "args": ["mcp-remote", "http://localhost:3000/mcp", "--transport", "http-only"],
      "env": { "NO_PROXY": "127.0.0.1,localhost" }
    }
  }
}

2. Cloudflare Worker (Local Dev)

bun x wrangler dev --local | cat

Create .dev.vars for local secrets:

BEARER_TOKEN=your_random_auth_token
TESSIE_ACCESS_TOKEN=your_tessie_token
TESSIE_VIN=your_vehicle_vin

Endpoint: http://127.0.0.1:8787/mcp

3. Cloudflare Worker (Deploy)

  1. Create KV namespace for session storage:
bun x wrangler kv:namespace create TOKENS

Output will show:

Add the following to your wrangler.toml:
[[kv_namespaces]]
binding = "TOKENS"
id = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
  1. Update wrangler.toml with your KV namespace ID:
[[kv_namespaces]]
binding = "TOKENS"
id = "your-kv-namespace-id-from-step-1"
  1. Set secrets:
# Generate a random token for client authentication
openssl rand -hex 32
bun x wrangler secret put BEARER_TOKEN
# Paste the generated token when prompted

# Tessie API credentials
bun x wrangler secret put TESSIE_ACCESS_TOKEN
# Paste your Tessie token when prompted

bun x wrangler secret put TESSIE_VIN
# Paste your VIN when prompted
  1. Deploy:
bun x wrangler deploy

Endpoint: https://<worker-name>.<account>.workers.dev/mcp

Client Configuration

Alice App

Add as MCP server with:

  • URL: https://your-worker.workers.dev/mcp
  • Type: streamable-http
  • Header: Authorization: Bearer <your-BEARER_TOKEN>

Claude Desktop / Cursor (Local Server)

{
  "mcpServers": {
    "tesla": {
      "command": "npx",
      "args": ["mcp-remote", "http://127.0.0.1:3000/mcp", "--transport", "http-only"],
      "env": { "NO_PROXY": "127.0.0.1,localhost" }
    }
  }
}

Claude Desktop / Cursor (Cloudflare Worker)

{
  "mcpServers": {
    "tesla": {
      "command": "npx",
      "args": ["mcp-remote", "https://your-worker.workers.dev/mcp", "--transport", "http-only"]
    }
  }
}

MCP Inspector (Quick Test)

bunx @modelcontextprotocol/inspector
# Connect to: http://localhost:3000/mcp (local) or https://your-worker.workers.dev/mcp (remote)

Tools

tesla_state

Get the current state of your Tesla vehicle.

// Input
{}

// Output
{
  display_name: string;           // Vehicle name
  battery_level: number;          // 0-100%
  battery_range_km: number;       // Estimated range in km
  charging: {
    state: string;                // "Disconnected", "Charging", "Complete", "Stopped"
    minutes_remaining: number | null;
    charge_limit: number;         // Charge limit %
  };
  location: {
    latitude: number;
    longitude: number;
    heading: number;              // 0-359°
    speed: number | null;         // km/h or null if parked
  };
  locked: boolean;
  sentry_mode: boolean;
  climate: {
    is_on: boolean;
    inside_temp: number;          // °C
    outside_temp: number;         // °C
    target_temp: number;          // °C
    is_defrosting: boolean;
  };
  doors: {
    front_left: boolean;          // true = open
    front_right: boolean;
    rear_left: boolean;
    rear_right: boolean;
    frunk: boolean;
    trunk: boolean;
    charge_port: boolean;
  };
  state: "online" | "asleep" | "offline";
  odometer_km: number;
  last_updated: string;           // ISO 8601
}

tesla_command

Execute commands on your Tesla vehicle.

// Input
{
  command: "lock" | "unlock" | "start_climate" | "stop_climate" |
           "set_temperature" | "start_defrost" | "stop_defrost" |
           "open_frunk" | "open_trunk" | "open_charge_port" |
           "close_charge_port" | "enable_sentry" | "disable_sentry" |
           "flash" | "honk" | "share";
  temperature?: number;           // Required for set_temperature (15-28°C)
  destination?: string;           // Required for share
  locale?: string;                // Optional for share (e.g., "en-US")
}

// Output
{
  success: boolean;
  command: string;
  message: string;
}

Commands Reference:

Command Description Parameters
lock Lock the vehicle —
unlock Unlock the vehicle —
start_climate Start climate control —
stop_climate Stop climate control —
set_temperature Set cabin temperature temperature (15-28°C)
start_defrost Turn on max defrost —
stop_defrost Turn off defrost —
open_frunk Open front trunk —
open_trunk Toggle rear trunk —
open_charge_port Open charge port door —
close_charge_port Close charge port door —
enable_sentry Enable sentry mode —
disable_sentry Disable sentry mode —
flash Flash the lights —
honk Honk the horn —
share Send destination to navigation destination, locale?

Examples

1. Get vehicle state

{
  "name": "tesla_state",
  "arguments": {}
}

2. Lock the car

{
  "name": "tesla_command",
  "arguments": {
    "command": "lock"
  }
}

3. Set temperature to 22°C

{
  "name": "tesla_command",
  "arguments": {
    "command": "set_temperature",
    "temperature": 22
  }
}

4. Start climate before leaving

{
  "name": "tesla_command",
  "arguments": {
    "command": "start_climate"
  }
}

5. Navigate to a destination

{
  "name": "tesla_command",
  "arguments": {
    "command": "share",
    "destination": "Golden Gate Bridge, San Francisco"
  }
}

Authentication Flow

┌─────────────────────────────────────────────────────────────────┐
│  Client (Alice App, Claude Desktop)                             │
│      │                                                          │
│      │ Authorization: Bearer <BEARER_TOKEN>                     │
│      ▼                                                          │
│  ┌─────────────────────────────────────────────────────────┐   │
│  │  Cloudflare Worker / Node.js Server                      │   │
│  │                                                          │   │
│  │  1. Validate BEARER_TOKEN (client auth)                  │   │
│  │  2. Use TESSIE_ACCESS_TOKEN (internal API key)           │   │
│  │                                                          │   │
│  │  env.TESSIE_ACCESS_TOKEN ──┐                             │   │
│  │  env.TESSIE_VIN ───────────┼──► TessieClient             │   │
│  │                            │         │                   │   │
│  │                            │         ▼                   │   │
│  │                            │   api.tessie.com            │   │
│  └─────────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────────┘

Key points:

  • BEARER_TOKEN: Random token you generate — authenticates clients to your MCP server
  • TESSIE_ACCESS_TOKEN: Your Tessie API key — used internally by the server
  • Clients never see your Tessie credentials

HTTP Endpoints

Endpoint Method Purpose
/mcp POST MCP JSON-RPC 2.0
/health GET Health check

Development

bun dev           # Start with hot reload
bun run typecheck # TypeScript check
bun run lint      # Lint code
bun run build     # Production build
bun start         # Run production

Architecture

src/
├── shared/
│   └── tools/
│       ├── tesla-state.ts      # Get vehicle state
│       └── tesla-command.ts    # Execute commands
├── services/
│   └── tessie.service.ts       # Tessie API client
├── schemas/
│   ├── commands.ts             # Command definitions
│   ├── outputs.ts              # Tool output schemas
│   └── tessie.ts               # Tessie API response schemas
├── config/
│   └── metadata.ts             # Server & tool descriptions
├── index.ts                    # Node.js entry
└── worker.ts                   # Workers entry

Environment Variables

Node.js (.env)

Variable Required Description
TESSIE_ACCESS_TOKEN ✓ Tessie API access token
TESSIE_VIN ✓ Tesla Vehicle VIN
BEARER_TOKEN ✓ Auth token for MCP clients
PORT Server port (default: 3000)
HOST Server host (default: 127.0.0.1)
AUTH_ENABLED Enable auth (default: true)
AUTH_STRATEGY bearer (default)

Cloudflare Workers (wrangler.toml + secrets)

wrangler.toml vars:

AUTH_ENABLED = "true"
AUTH_STRATEGY = "bearer"

Secrets (set via wrangler secret put):

  • BEARER_TOKEN — Random auth token for clients
  • TESSIE_ACCESS_TOKEN — Tessie API access token
  • TESSIE_VIN — Your vehicle's VIN

KV Namespace:

[[kv_namespaces]]
binding = "TOKENS"
id = "your-kv-namespace-id"

Troubleshooting

Issue Solution
401 Unauthorized Check BEARER_TOKEN is set and client sends Authorization: Bearer <token>
"TESSIE_ACCESS_TOKEN not configured" Set secret: wrangler secret put TESSIE_ACCESS_TOKEN
"TESSIE_VIN not configured" Set secret: wrangler secret put TESSIE_VIN
"Tessie API error" Verify TESSIE_ACCESS_TOKEN is valid at developer.tessie.com
Vehicle not found Check TESSIE_VIN is correct (17 characters)
Vehicle offline Vehicle may be in deep sleep. Commands will wake it (takes ~30s)
Command timeout Tessie waits up to 90s for vehicle wake. Try again.
KV namespace error Run wrangler kv:namespace create TOKENS and update wrangler.toml
"ReadableStream is not defined" Node.js version too old (needs 18+). Use full path to newer node.
"spawn bunx ENOENT" Claude Desktop can't find bunx. Use npx instead.

Debugging

Test with MCP Inspector:

bunx @modelcontextprotocol/inspector
# Connect to your endpoint and test tools

Check Worker logs:

wrangler tail

License

MIT

MCP Server · Populars

MCP Server · New

    reticlehq

    reticle

    AI agents can generate code, but still struggle to understand what they build. Reticle brings Jev-style machine-native runtime perception to web & desktop applications.

    Community reticlehq
    tubeagentkit

    YouTube MCP Server: Transcripts, Search, Channels & Playlists

    YouTube MCP server: get YouTube video transcripts, search YouTube, browse channels and playlists from Claude, ChatGPT, Cursor and VS Code. Remote, free tier.

    Community tubeagentkit
    lrvx

    lrvx

    Framework for building trading systems

    Community lrvx
    koda-ai-studio

    agent-bom

    Open security scanner and self-hosted control plane for AI, MCP, and cloud. One evidence model — run scans in your environment, centralize findings, govern in your VPC.

    Community koda-ai-studio
    DenRakEiw

    Scumble

    Free, open-source desktop editor for AI inpainting: select, prompt, get a colour-matched layer. FLUX 3 Image, FLUX.2, GPT Image, Nano Banana, Seedream and Qwen through your API key or your own ComfyUI. Layers, PSD export, MCP.

    Community DenRakEiw