Automatic
Anonymization
with LibreChat

Original:

"My name is John Doe, email: john@example.com"

Anonymized:

"My name is [[Person-abc123]], email: [[Email-def456]]"

German, English, French, Spanish, Italian

• API issues: https://app.anymize.ai
• LibreChat documentation: https://docs.librechat.ai
• anymize website: https://anymize.ai

• Never share API keys publicly
• Data is processed securely according to GDPR standards
• Only the original session can de-anonymize data
• No data storage – processing happens in real-time

Easy setup, use in chat interfaces, no coding required

Get API key at https://app.anymize.ai. Add to librechat.yaml:

mcpServers:
  anymize:
    type: streamable-http
    url: 'https://mcp.anymize.ai/mcp'
    timeout: 60000
    chatMenu: true
    startup: true
    headers:
      x-anymize-api-key: anymize_yourkey

Replace anymize_yourkey with your actual API key. Restart LibreChat afterwards.

• Select a Tools-compatible endpoint (e.g. OpenAI, Anthropic, Google)
• Open the MCP Server dropdown below the text input
• Select Anymize
• Start chatting – the AI can now anonymize and de-anonymize text

• Open the Agent Builder
• Click Add Tools
• Save the agent

Advanced file processing (PDF/image OCR), custom configurations

• LibreChat with structured tools support
• Node.js 16+
• anymize API key from https://app.anymize.ai

Download Anymize.js from Google Drive. Place the file at:

api/app/clients/tools/structured/Anymize.js

Edit api/app/clients/tools/manifest.json – add:

{
  "name": "Anymize",
  "pluginKey": "anymize",
  "description": "GDPR-compliant text anonymization and deanonymization.",
  "icon": "https://res.cloudinary.com/.../anymize-favico-b_tpvtwh.svg",
  "authConfig": [
    {
      "authField": "ANYMIZE_API_KEY",
      "label": "Anymize API Key",
      "description": "Get your API key from Anymize Dashboard"
    }
  ]
}

Edit api/app/clients/tools/index.js:

const Anymize = require('./structured/Anymize');

module.exports = {
  // ... existing exports
  Anymize,
};

Edit api/app/clients/tools/util/handleTools.js – add import:

const { Anymize } = require('../');

Add anymize to the customConstructors object:

anymize: async (toolContextMap) => {
  const authFields = getAuthFields('anymize');
  const authValues = await loadAuthValues({ userId: user, authFields });

  const contextFiles = options.tool_resources?.[EToolResources.context]?.files ?? [];
  const messageAttachments = options.req?.body?.files ?? [];
  const allFiles = [...contextFiles];
  const seenFileIds = new Set(contextFiles.map(f => f?.file_id).filter(Boolean));

  for (const file of messageAttachments) {
    if (file?.file_id && !seenFileIds.has(file.file_id)) {
      allFiles.push(file);
      seenFileIds.add(file.file_id);
    }
  }

  if (allFiles.length > 0) {
    let toolContext = 'Context files available for anonymization:';
    for (const file of allFiles) {
      if (file) toolContext += `\n\t- ${file.filename} (${file.file_id})`;
    }
    toolContextMap.anymize = toolContext;
  }

  return new Anymize({
    ...authValues,
    userId: user,
    req: options.req,
    contextFiles: allFiles,
  });
},

Add the API key to your .env file:

ANYMIZE_API_KEY=your_api_key_here

Restart LibreChat.

• Anonymize text: "Anonymize this: My name is John Doe"
• Anonymize files: Upload a PDF or image and say "Anonymize this document"
• De-anonymize: "Deanonymize: [[Person-abc123]]"
• Supported file formats: PDF, PNG, JPG (processed via OCR)

Server not showing:

• Check the API key in librechat.yaml
• Restart LibreChat
• Make sure you selected a Tools-compatible model

Connection timeout:

• Check your internet connection
• Check the status of the Anymize service

Tool not showing:

• Make sure all files are in the correct directories
• Check JSON syntax in manifest.json
• Clear browser cache

"Missing API Key" error:

• Check that ANYMIZE_API_KEY=your_key is in the .env file
• Make sure there are no spaces around the "="
• Restart the backend after changes

"No files available" error:

• Upload a file first
• Wait for the upload to complete
• Supported formats: PDF, PNG, JPG