omni-lpr — Claude Code MCP server for license plate recognition
omni-lpr — a self-hosted server that finds and reads license plates in images. It talks to agents over MCP and to any app over REST, so your main codebase stays free of ML dependencies.
MIT ★ 27 updated 2026-07-05
Open on GitHubWhat omni-lpr does
omni-lpr detects plates in a full photo and reads the text from pre-cropped plate images. Pictures can arrive as Base64, file upload, a local path or a URL. The tools list available detector and OCR models as well.
Run it with pip install omni-lpr, then start it with the omni-lpr command. It listens on http://127.0.0.1:8000 by default. The MCP endpoint sits at http://127.0.0.1:8000/mcp/ over streamable HTTP, and Swagger UI is at http://127.0.0.1:8000/api/v1/apidoc/swagger. MCP Inspector works for poking at the tools before wiring an agent in.
Python 3.10 to 3.13 is required. Docker images cover plain CPU via ONNX, Intel chips via OpenVINO and NVIDIA GPUs via CUDA. The project is in early development, so API changes that break things are likely — pin a version if you build on it.
What it helps with
- detect every plate in a full image and read the text
- recognize text from an already cropped plate picture
- give an AI agent license plate recognition through MCP tools
- keep plate recognition as a separate service instead of dragging Python and ML deps into the main app
- pick hardware acceleration: ONNX for CPU, OpenVINO for Intel, CUDA for NVIDIA
Example prompts for the agent
- Read the plate from https://www.olavsplates.com/foto_n/n_cx11111.jpg
- Find and recognize all license plates in this image
- List the available detection and OCR models
How to install omni-lpr
1Template address
The server address is a template: replace {…} or […] with your own values
2Platform CLI: remote server
claude mcp add --transport http omni-lpr https://{baseUrl}/mcp/1Template address
The server address is a template: replace {…} or […] with your own values
2Platform CLI: remote server
codex mcp add omni-lpr --url https://{baseUrl}/mcp/1Template address
The server address is a template: replace {…} or […] with your own values
2Config file
{
"mcpServers": {
"omni-lpr": {
"url": "https://{baseUrl}/mcp/"
}
}
}1Template address
The server address is a template: replace {…} or […] with your own values
2Config file
{
"servers": {
"omni-lpr": {
"type": "http",
"url": "https://{baseUrl}/mcp/"
}
}
}1Template address
The server address is a template: replace {…} or […] with your own values
2Platform CLI: remote server
gemini mcp add -t http omni-lpr https://{baseUrl}/mcp/Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about omni-lpr
How do I install omni-lpr in Claude Code?
Start the server first: pip install omni-lpr, then run omni-lpr. Add it to Claude Code with claude mcp add --transport http omni-lpr https://{baseUrl}/mcp/, replacing {baseUrl} with the actual address.
What does omni-lpr do?
It finds license plates in photos and reads the text from them. You can feed it a full image or an already cropped plate, and it returns results through REST or MCP.
Does it need a GPU?
No. It runs on plain CPUs through ONNX, with separate Docker images for Intel OpenVINO and NVIDIA CUDA if you want hardware acceleration.