FileAI

The fileAI MCP Server provides an integration layer that lets you upload files, run OCR, classify documents, and extract structured data, all through the MCP protocol. It enables AI models to work with your documents programmatically, handling processing asynchronously and giving you control over extraction schemas.

How to use

Connect your MCP client to the fileAI MCP Server to perform end-to-end file processing. You can start by using a remote HTTP MCP endpoint for server-side processing or run a local MCP client that spawns the server process as a stdio app. Once connected, upload files, trigger OCR, have the system classify documents, and request structured data according to your defined schemas. All steps can be tracked asynchronously so you can monitor progress and fetch results when ready.

Key capabilities you can leverage include uploading files, OCR, document classification, and structured data extraction. Define and manage extraction schemas to control what data you retrieve, and use asynchronous processing to decouple submission from result retrieval.

How to install

Prerequisites you need before configuring the MCP server include a fileAI account, a fileAI API Key, verified AI schemas if you want deterministic responses, an MCP client, and npx installed in your environment.

Install Node.js and npm if you don’t have them already. Then proceed to configure the MCP server connection using either the remote HTTP method or a local stdio command as shown in the configuration example.

Create the MCP configuration with your API key and endpoints, then verify the setup by ensuring the MCP tools are enabled in your client.

Additional notes

The fileAI MCP Server can be connected via an HTTP MCP endpoint or by running a local stdio process that uses npx to invoke the remote MCP resources. For Claude integration, add the server URL to Claude's integrations with your API key, keeping in mind Claude’s policies on PII. If you encounter issues, contact support@file.ai for assistance.

Configuration and examples

{
  "mcpServers": {
    "FileAIHttp": {
      "type": "http",
      "name": "fileai_http",
      "url": "https://mcp.file.ai/mcp",
      "args": []
    },
    "FileAIFullLocal": {
      "type": "stdio",
      "name": "fileai_local_mcp",
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.file.ai",
        "--header",
        "accept: application/json",
        "--header",
        "x-api-key:${API_KEY}"
      ],
      "env": {
        "API_KEY": "<YOUR_FILEAI_API_KEY>"
      }
    },
    "FileAIClaude": {
      "type": "http",
      "name": "fileai_claude_integration",
      "url": "https://mcp.file.ai/sse?x-api-key=Bearer YOUR_API_API_KEY_HERE",
      "args": []
    }
  }
}

Security and environment

Keep your API keys secure. Use environment variables to inject keys into MCP configurations and avoid embedding secrets directly in code blocks. If you use Claude, ensure your use case complies with Claude’s terms of service regarding PII.

Available tools

Upload and process files

Submit files for processing, OCR, classification, and extraction tasks through MCP.

OCR processing

Perform optical character recognition to extract text from documents.

Document classification

Classify documents into predefined categories to route processing and schemas.

Schema-driven extraction

Define and manage extraction schemas to control the structure of returned data.

Asynchronous job tracking

Submit jobs and monitor their status until results are available.