vin-lookup-mcp

vin-lookup-mcp is an MCP server for NHTSA vPIC lookups. It exposes two tools:

• decode_vin for 17-character VIN decoding
• get_canadian_vehicle_specifications for Canadian-market vehicle specs

Example prompt

Use only the registered MCP tool `decode_vin` for each VIN below. Do not inspect local files, search my home directory, or invoke `python`/`uv` directly. Decode each VIN with the MCP tool and give me a short summary for each one.

5YJ3E1EB5KF123456
1N4AL21E87C118393
1P8ZA1279SZ215470
1HGCM82633A004352
1FTFW1EF1EKE12345
JHMFA16586S012345
2HGFG3B59FH512345
3VWFE21C04M000001
1G1JC5244T7258391
JTDKB20U093123456

Free API

This project uses NHTSA's public vPIC API:

• API docs: https://vpic.nhtsa.dot.gov/api/Home
• Decoder UI: https://vpic.nhtsa.dot.gov/decoder/

No API key is required.

MCP tools

The server exposes two MCP tools:

• decode_vin(vin: str)
• get_canadian_vehicle_specifications(year: int, make: str, model?: str, units?: "Metric" | "US")

decode_vin:

• normalizes the VIN
• validates that it is a 17-character VIN
• calls DecodeVinValuesExtended
• returns:
  • vin
  • summary
  • raw
  • api_error_code
  • api_error_text

get_canadian_vehicle_specifications:

• calls NHTSA's GetCanadianVehicleSpecifications
• accepts year and make, with optional model and units
• returns:
  • search_criteria
  • count
  • message
  • summary
  • results

Run locally

Start the MCP stdio server:

uv run python -m vin_lookup_mcp.server

Test without MCP

You can decode a VIN directly from the command line:

uv run python -m vin_lookup_mcp.server --decode 1HGCM82633A004352

For just the compact summary:

uv run python -m vin_lookup_mcp.server --decode 1HGCM82633A004352 --summary-only

You can also test the Canadian specifications endpoint directly:

uv run python -m vin_lookup_mcp.server \
  --canadian-specs-year 2011 \
  --canadian-specs-make Acura

For just the compact summary:

uv run python -m vin_lookup_mcp.server \
  --canadian-specs-year 2011 \
  --canadian-specs-make Acura \
  --summary-only

For a quick end-to-end CLI smoke test, run:

bash scripts/smoke_test.sh

Project structure

.
├── pyproject.toml
├── README.md
├── scripts
│   └── smoke_test.sh
└── vin_lookup_mcp
    ├── __init__.py
    └── server.py

Notes

• vPIC is strongest for vehicles intended for sale or import into the United States.
• Canadian specifications come from the dedicated vPIC GetCanadianVehicleSpecifications endpoint.
• NHTSA notes that VIN decoding is for model years 1981 and newer.
• The MCP server uses the official Python MCP SDK via FastMCP.

Why I built this

I was shopping for cars and wanted to talk to ChatGPT about them by pasting the VIN number rather than the full model information.

At my job we use a lot of MCP servers so I wanted to learn how to build my own. I only used Codex for this.