MCP Mistral OCR is an MCP server that provides optical character recognition capabilities by connecting Model Context Protocol clients directly to the Mistral AI OCR API. Software developers, data analysts, and researchers use this server to automate document digitization without leaving their AI workflow. It enables client applications like Claude Desktop to parse text, structured layouts, and embedded data from local disk storage or remote web addresses. Supported media include standard document formats like PDF as well as image files like JPG, PNG, GIF, and WebP. The server processes local files placed within a mapped directory or pulls documents directly from specified URLs. Processing results are systematically written to timestamped JSON files stored in a designated output directory. By offloading OCR heavy lifting to Mistral AI's cloud infrastructure, the server handles documents up to 50MB in size and up to 1000 pages per document, giving AI assistants access to unstructured visual files.
Category: Design, Media & Creative
Tags: computer-vision, image-processing, mistral, ocr, pdf
Follow these steps to install and configure MCP Mistral OCR for Claude Desktop using Docker: 1. Build the Docker container image locally: bash docker build -t mcp-mistral-ocr . 2. Open your claude_desktop_config.json configuration file. 3. Add mistral-ocr under the mcpServers section: json { "mcpServers": { "mistral-ocr": { "command": "docker", "args": [ "run", "-i", "--rm", "-e", "MISTRAL_API_KEY", "-e", "OCR_DIR", "-v", "/path/to/local/files:/data/ocr", "mcp-mistral-ocr:latest" ], "env": { "MISTRAL_API_KEY": "<YOUR_MISTRAL_API_KEY>", "OCR_DIR": "/path/to/local/files" } } } } 4. Replace <YOUR_MISTRAL_API_KEY> with your actual Mistral API key and /path/to/local/files with your local document folder. 5. Restart Claude Desktop to load the server.
Part of MCP Servers
MCP Mistral OCR is a Model Context Protocol server that connects Claude Desktop and other MCP-compatible clients to Mistral AI's OCR API. It enables users to run optical character recognition on local files and remote web links.
The server supports common image formats including JPG, JPEG, PNG, GIF, and WebP, as well as PDF documents. File processing is subject to Mistral API limitations, which cap document size at 50MB and up to 1000 pages.
OCR output is automatically exported as structured JSON files inside the output folder within your configured OCR directory. Each output file is timestamped with the date and time of execution alongside the original document name.
The server exposes two main tools: process_local_file, which reads and processes documents stored in the configured OCR directory, and process_url_file, which fetches and analyzes documents from a web URL using a specified file type argument.