A question for the AI age

What if every AI knew when your car was lying to you?

MED MCP is the first automotive Model Context Protocol server. Real repair data, real maintenance schedules, real shop-floor patterns — plugged directly into the AI you already use.

Install on Smithery → See how it works

The problem

Ask any chatbot about your car. You'll get advice that's generic, vague, and often wrong.

"Sounds like it could be a few things — you should take it to a mechanic."

Generic AI, on every car question, ever. Not good enough.

Introducing

MED MCP.

The first automotive Model Context Protocol server. Real repair data, real maintenance schedules, real shop-floor patterns — plugged directly into the AI you already use.

1st

Automotive MCP server

20yrs

Mechanic experience built-in

API keys · No external services

Four tools

Everything an AI needs to give you real car advice.

check_repair_estimate( )

Compares a quote against national price ranges. Returns in-range / high / low — plus questions to ask the shop.

get_maintenance_schedule( )

What's due now, what's due in the next 10K, common upsells, and known issues for the make and year.

should_i_fix_or_replace( )

Repair-to-value ratio with mileage, age, and history factored in. A scored, defensible recommendation.

check_shop_red_flags( )

Reads shop interactions for 7 known red-flag patterns. Returns matches, positive signals, and what to do next.

Three audiences

Built for the whole conversation.

For drivers

Anyone with a car
and a question.

Walk into the shop knowing what fair looks like, what's actually due, and which questions make a service writer back off.

For AI assistants

Claude, ChatGPT,
and friends.

Drop-in MCP server. Your assistant stops guessing about cars and starts citing repair ranges, intervals, and red-flag patterns.

For mechanics

Honest shops with
nothing to hide.

A neutral reference customers can check against. The good shops want this — it ends the arguments before they start.

How it's different

Built by a mechanic, not a tech bro.

20 years under the hood — military and civilian. Sister product to the Driver's Survival Guide, the MED newsletter, and the advisory service. The knowledge is the product.

Architecture

No API keys. No external services. No database.
All data ships in editable JSON.

Sourced from

RepairPal · AAA · AutoLeap
Cross-checked against shop-floor reality.

35 repair types
6 vehicle models
7 red-flag patterns
5 maintenance intervals
20+ shop questions

One link, no config

One link.
One install.

Open Smithery, click install, point your AI client at MED MCP. That's it — no keys, no config files, no waiting list.

$ npx -y @smithery/cli install @zimconsulting-ops/med-vehicle-intelligence --client claude

Smithery — recommended smithery.ai/server/@zimconsulting-ops/med-vehicle-intelligence

GitHub — source & issues github.com/zimconsulting-ops/med-vehicle-intelligence

Install on Smithery → View on GitHub

Manual install (advanced)

Not on a Smithery-supported client — or just want to skip the proxy? Install the Python package directly and point your MCP config at it.

1. Install the package

$ pip install med-vehicle-intelligence

2. Add to your MCP config

Claude Desktop config lives at %APPDATA%\Claude\claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS. Cursor and Cline accept the same shape in their own MCP config files.

{
  "mcpServers": {
    "med-vehicle-intelligence": {
      "command": "med-vehicle-intelligence"
    }
  }
}

Restart your client. The four MED tools appear automatically — same data, no Smithery in the path.