MCP SERVER

Add Privacy Tools to Your Claude Agent Loop

Connect our MCP server. 7 privacy tools auto-available in Claude. Real-time PII detection and redaction in agentic workflows. Zero data cloud uploads.

7 Tools in Your Workflow

analyze_text

Scan text for PII. Returns entity type, confidence, position. Use in RAG pipeline before LLM context.

anonymize_text

Redact PII from text. Methods: mask, hash, encrypt, remove. Agent decides which method per entity type.

scan_file

Upload PDF/DOCX/XLSX. Detect all PII. Return suggestions. Pre-redaction QA step in workflow.

batch_anonymize

Process 5000+ texts in parallel. Returns all results in array. Useful for bulk customer data.

get_entities

List all 285+ PII entity types. Filter by language or domain. Agent discovers available entities.

estimate_cost

Preview token cost before processing. Agent checks budget, optimizes batch sizes accordingly.

Setup Instructions (5 minutes)

Step 1: Get API Key

Sign up at anonym.legal. Copy your API key from settings. Keep it secret (treat as password).

Step 2: Configure claude_desktop_config.json

{
  "mcpServers": {
    "anonym": {
      "command": "npx",
      "args": ["@anonym-legal/mcp-server"],
      "env": {
        "ANONYM_API_KEY": "your_api_key_here"
      }
    }
  }
}

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Step 3: Restart Claude

Quit Claude Desktop completely. Relaunch. Check settings → "Privacy Tools" should appear. Green checkmark = connected.

Step 4: Test in Chat

Try: "Can you scan this text for PII: My SSN is 123-45-6789?"
Claude calls analyze_text tool. Returns results instantly.

Agentic Workflow Example

# Claude's reasoning loop:
# 1. User: "Process these customer records and summarize insights"
# 2. Claude calls scan_file (uploads CSV with customer emails/SSNs)
# 3. Tool returns: 523 PII entities found (285 emails, 238 SSNs)
# 4. Claude calls batch_anonymize (redacts all SSNs, keeps emails)
# 5. Claude processes redacted data, generates summary
# 6. User gets insights WITHOUT raw PII exposure
# 7. Audit log: "batch_anonymize (523 SSNs masked)" logged

Enterprise Features

🔐 Bearer Token Auth

API key never exposed in code. Sent as Bearer header. Session tokens <55min TTL.

📊 Audit Log

Every tool call logged: timestamp, user, entities, method. SIEM integration available.

🔄 Rate Limiting

Fair usage: 1000 requests/min per key. Burst to 5000. Contact us for enterprise limits.

🌍 Scalable Infrastructure

Distributed endpoints. Sub-200ms latency. Auto-failover on regional outage.

⚡ Parallelism

Process multiple requests concurrently. Agent loop stays responsive.

📈 Production Ready

Enterprise-grade performance. No pre-provisioning needed.

Troubleshooting

1. Check API key in config (no spaces/quotes). 2. Restart Claude completely (not refresh). 3. Check console: Help → "View Logs" for errors.

API key expired or revoked. Regenerate at dashboard. Update config. Restart Claude.

Check network latency to closest region. Contact support for regional routing optimization.

Watch the API In Action

See PII detection and anonymization via REST API and MCP Server

Connect Privacy Tools Now

Get API key. 5 minutes to setup. Add 7 privacy tools to Claude agent loop.

Get Started

Also from anonym.legal

Enterprise Deployment → Anonymization Methods → Compliance Presets → EU Entity Coverage →

Frequently Asked Questions

Add the anonym.legal MCP server to your Cursor settings (Settings → MCP → Add Server). Configuration: server URL https://anonym.legal/mcp, Bearer token authentication. Cursor AI can then use 7 PII protection tools: analyze, anonymize, deanonymize, supported-entities, health, language-detect, and entity-count.

Yes. Any VS Code extension that supports MCP (Continue.dev, Cline) can connect to anonym.legal's MCP Server. Add the server configuration to your MCP settings file and the AI assistant gains PII detection capabilities.

analyze (detect PII), anonymize (redact PII), deanonymize (reverse encryption), supported-entities (list detectable types), health (server status), language-detect (identify text language), entity-count (count PII in text).