vin-decode-mcp

# vin-decode-mcp **Decode VINs and query vehicle data from a curated NHTSA vPIC database — powered by the Model Context Protocol.** A standalone, offline-capable MCP server for LLMs to decode Vehicle Identification Numbers (VINs) and look up makes, models, and vehicle specifications using data from [NHTSA's vPIC](https://vpic.nhtsa.dot.gov/). ``` pip install vin-decode-mcp vin-decode-mcp # Start the MCP server ``` ## Why? - **Offline**: Works without internet access. The curated SQLite database (~4.5 MB) is self-contained. - **No rate limits**: Unlike calling the vPIC API directly, local queries are unlimited. - **Fast**: Pattern matching against the SQLite database takes microseconds. - **LLM-native**: Tools with rich docstrings, schema resources, and structured JSON output. - **Open data**: NHTSA vPIC is US government open data — free, no API key required. ## Data Coverage - US-market vehicles, model year **1981 and forward** - **536 makes**, **9,284 models**, **88,267 VIN patterns** (2026-08 vintage) - Passenger Cars, Trucks, MPVs, Motorcycles, Off-Road Vehicles - Excludes: Buses, Trailers, Low-Speed Vehicles, Incomplete Vehicles > **Specifications only** — this database does not include title, accident, odometer, > or theft history (those require NMVTIS/commercial data sources). ## Quick Start ### Installation ```bash pip install vin-decode-mcp ``` Or from source: ```bash git clone https://github.com/<org>/vin-decode-mcp.git cd vin-decode-mcp pip install -e . ``` ### Running ```bash # Default: stdio transport (for Claude Desktop, Cursor, etc.) vin-decode-mcp # HTTP transport vin-decode-mcp --transport http --port 8765 ``` ### Using with Claude Desktop Create a dedicated venv so the binary lands where you can reference it: ```bash python3 -m venv ~/.local/venvs/vin-decode source ~/.local/venvs/vin-decode/bin/activate pip install vin-decode-mcp deactivate ``` Add to `~/.config/claude-desktop/config.json` (or `~/Library/Application Support/cla…