MCP & AI Integration | anonymize.solutions

Streamable HTTP

Bearer Auth

7 MCP Tools

E2E Encryption

Protocol

Streamable HTTP Transport

The MCP server implements Streamable HTTP transport as defined in the MCP March 2025 specification. Two transport modes are supported for different client environments.

stdio

Claude Desktop

Standard input/output transport for direct integration with Claude Desktop. The MCP server runs as a local process and communicates via stdio streams.

Local process execution
Claude Desktop native integration
stdio stream communication

HTTP

Cursor & VS Code

HTTP transport for IDE-based clients. The MCP server exposes an HTTP endpoint that Cursor IDE and VS Code extensions connect to.

HTTP endpoint communication
Cursor IDE integration
VS Code extension support
MCP-compatible tool support

Tools

7 MCP Tools

Each tool is exposed as an MCP function that AI assistants can invoke directly. Token costs are deducted from the account balance per operation.

anonym_legal_analyze_text

Analyze Text for PII

Detect personally identifiable information in text. Returns entity positions, types, and confidence scores without modifying the original text.

2 base + 0.5/KB + 0.2/entity filter

anonym_legal_anonymize_text

Anonymize Text

Replace detected PII with anonymized placeholders. Supports configurable operators (replace, mask, hash) per entity type.

3 base + 0.5/KB + 0.3/unique entity

anonym_legal_detokenize_text

Detokenize Text

Restore original values from anonymized tokens using session-stored or client-provided token mappings.

2 base + 0.3/token

anonym_legal_get_balance

Check Token Balance

Query the current token balance for the authenticated account. Returns remaining tokens and usage statistics.

Free

anonym_legal_estimate_cost

Estimate Cost

Preview the token cost of an operation before executing it. Useful for budget-aware AI workflows.

Free

anonym_legal_list_sessions

List Sessions

View active anonymization sessions with their metadata. Sessions store token mappings for detokenization.

Free

anonym_legal_delete_session

Delete Session

Permanently delete an anonymization session and its associated token mappings from server storage.

Free

Setup

Configuration

Add the MCP server to your AI client configuration. The package is distributed via npm and runs with npx.

MCP Client Configuration

{
  "mcpServers": {
    "anonym-legal": {
      "command": "npx",
      "args": ["-y", "anonym-legal-mcp"],
      "env": {
        "ANONYM_API_KEY": "your-api-key-here"
      }
    }
  }
}

Package: npx -y anonym-legal-mcp (npm)

Parameters

Detection Parameters

Customize PII detection behavior through MCP tool parameters. These parameters apply to the analyze and anonymize tools.

Parameter	Description
language	Language of the input text for language-specific detection models
entities	Specific entity types to detect (e.g., PERSON, PHONE_NUMBER, EMAIL_ADDRESS)
entity_groups	Predefined groups of entity types for common use cases
score_threshold	Minimum confidence score for entity detection (0.0 to 1.0)
preset	Predefined detection configuration for common scenarios
ad_hoc_recognizers	Custom regex-based recognizers defined inline for domain-specific patterns
output	Output format configuration for results
response_format	Structure of the API response (detailed or compact)

Customization

Custom Operators via MCP

Define per-entity-type anonymization operators. Each entity type can use a different strategy: replace with a placeholder, mask with characters, or hash the value.

{
  "operators": {
    "PERSON": {
      "type": "replace",
      "new_value": "[REDACTED_NAME]"
    },
    "PHONE_NUMBER": {
      "type": "mask",
      "masking_char": "*",
      "chars_to_mask": "6",
      "from_end": "true"
    },
    "EMAIL_ADDRESS": {
      "type": "hash",
      "hash_type": "sha256"
    }
  }
}

End-to-End

E2E Encryption via MCP

End-to-end encryption splits the anonymization process between server and client. The server detects PII positions; the client extracts, encrypts, and stores the original values locally. The server never sees the original PII text.

1. Server Detects Positions

The MCP server analyzes text and returns entity positions and types without storing the original text.

2. Client Extracts & Encrypts

The MCP client extracts PII from the original text using positions and encrypts values locally.

3. Client Stores Locally

Encrypted token mappings are stored on the client side. The server has no access to decryption keys.

Client-Side Utility Functions

buildE2ETokenMappings()

Build encrypted token mappings from server-detected positions and the original text.

detokenizeLocally()

Restore original text using locally stored encrypted token mappings.

serializeMappings()

Serialize token mappings for persistent storage or transfer between sessions.

deserializeMappings()

Deserialize previously stored token mappings back into usable form.

Specifications

Technical Specifications

Authentication

Bearer Token

Rate Limit

100 req/min

Max Text Size

100 KB

Max Request Body

1 MB

Session TTL

24 hours

Persistent Sessions

30 days

Compatibility

Compatible Clients

The MCP server is compatible with any MCP-capable AI client. Tested and supported on the following platforms.

Claude Desktop

stdio transport

Cursor IDE

HTTP transport

VS Code

HTTP transport

MCP-Compatible Tools

Any MCP client

Integrate AI-Powered Anonymization

Request access to the MCP Server for your Enterprise deployment. Available on Pro and Business plans.

Request a Quote View Enterprise Plans