Mistral OCR

Model Context Protocol (MCP) Server for Mistral OCR API

python

37

GitHub Stars

python

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "everaldo-mcp-mistral-ocr": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "MISTRAL_API_KEY",
        "-e",
        "OCR_DIR",
        "-v",
        "C:/path/to/your/files:/data/ocr",
        "mcp-mistral-ocr:latest"
      ],
      "env": {
        "OCR_DIR": "C:/path/to/your/files",
        "MISTRAL_API_KEY": "<YOUR_MISTRAL_API_KEY>"
      }
    }
  }
}

You can run OCR processing for local images and PDFs or files fetched from URLs using the Mistral OCR MCP Server. It supports a wide range of input formats, saves outputs as timestamped JSON, and can run inside Docker or via local development workflows for easy integration with your MCP client workflows.

How to use

To use this MCP server, run it in a way that matches your environment, either through Docker or a local development setup. You can submit two kinds of tasks: processing a file already available in your OCR input directory, and processing a file fetched from a URL with an explicit file type. The server will store the OCR results as JSON files with timestamps inside the configured OCR directory.

Common workflows include placing local files (images or PDFs) into your OCR_DIR, then invoking the process_local_file tool with the filename. For remote content, call process_url_file with the URL and the file type (image or pdf) to ensure correct handling. The output files are saved under the output area in OCR_DIR with names that include the input filename and a timestamp, making it straightforward to track results.

How the outputs are organized

OCR results are saved as JSON files inside the output directory located within your OCR_DIR. For local inputs, outputs use the original filename with a timestamp, for example document.pdf_20240101_123456.json. For URL inputs, the filename is derived from the URL when available, otherwise a fallback like url_document_<timestamp>.json is used.

Usage tips

Ensure your MISTRAL_API_KEY is provided when running the server so requests to Mistral OCR can be authenticated.
Keep OCR_DIR mapped to a persistent path on your host to retain inputs and outputs across restarts.

Available tools

process_local_file

Process a file from the configured OCR_DIR directory by providing the filename.

process_url_file

Process a file from a URL by providing the URL and the file type (image or pdf).