Content
# MED Vehicle Intelligence MCP Server
Expert car repair and maintenance advisory tools from a 20-year mechanic veteran — available as an MCP server for AI assistants.
When someone asks Claude, Cursor, or any MCP-compatible assistant about car repairs, this server provides structured, expert-level answers based on real industry data.
## What It Does
Four tools that answer the questions every car owner asks:
### `check_repair_estimate`
*"Is $800 fair for front brakes on a 2020 Camry?"*
Compares a repair quote against national price ranges for independent shops and dealers. Returns whether the price is in range, high, or low — plus questions to ask the shop.
### `get_maintenance_schedule`
*"What maintenance does a 2018 F-150 need at 75,000 miles?"*
Returns what's due now, what's coming in the next 10,000 miles, common upsells to question, and known issues for your specific vehicle.
### `should_i_fix_or_replace`
*"Should I fix my car or buy a new one? Repair is $3,000, car is worth $8,000."*
Calculates repair-to-value ratio, factors in mileage, age, and repair history, then gives a scored recommendation. Includes the real cost of replacement for comparison.
### `check_shop_red_flags`
*"The shop said they won't know the cost until they get in there and want me to authorize the work."*
Analyzes a shop interaction for 7 common red flag patterns. Returns matched flags, positive signals, an overall assessment, and what to do next.
## Knowledge Base
All data lives in editable JSON files — no hardcoded values:
- **35 repair types** with independent and dealer price ranges (sourced from RepairPal, AAA, AutoLeap)
- **6 vehicle models** with known issues and typical costs (Camry, F-150, Civic, Silverado, RAV4, CR-V)
- **7 shop red flag patterns** with detection keywords and remediation advice
- **5 maintenance intervals** with priority levels and common upsell warnings
- **20+ shop questions** organized by context (estimate, diagnosis, repair vs replace)
## Install
```bash
pip install med-vehicle-intelligence
```
### Claude Desktop
Add to your `claude_desktop_config.json`:
```json
{
"mcpServers": {
"med-vehicle-intelligence": {
"command": "python",
"args": ["-m", "med_vehicle_intelligence"]
}
}
}
```
### Via Smithery
```bash
npx -y @smithery/cli install @zimconsulting-ops/med-vehicle-intelligence --client claude
```
### From source
```bash
git clone https://github.com/zimconsulting-ops/med-vehicle-intelligence.git
cd med-vehicle-intelligence
pip install -e .
```
## Run from terminal
Three equivalent ways to launch the server. The server reads JSON-RPC over stdio — once started, the cursor will sit blank waiting for input. Press `Ctrl+C` to exit.
```bash
# Preferred — entry-point script (works when Python's Scripts folder is on PATH)
med-vehicle-intelligence
# Package module (works anywhere Python is installed; recommended for Claude Desktop)
python -m med_vehicle_intelligence
# Server module (explicit submodule path; equivalent to the above)
python -m med_vehicle_intelligence.server
```
## Requirements
- Python 3.11+
- `mcp` package (installed automatically)
No API keys. No external services. No database. The knowledge is the product.
## About
Built by [My Everyday Driver](https://myeverydaydriver.com) — an automotive intelligence platform for everyday vehicle owners. Founded by a veteran with 20 years of military and civilian automotive experience.
The same expertise behind the Driver's Survival Guide, the MED newsletter, and the advisory service — now available to any AI assistant.
## License
MIT
