MCP server

Add Siper's demo MCP server to tools like Claude Code so your text is masked before it reaches the model.

You can use Siper’s masking from inside AI tools such as Claude Code. The Model Context Protocol (MCP) is an open standard that defines how such tools call external services as tools; Siper publishes a demo server that speaks this standard:

https://siper.gurubase.ai/api/mcp

The server offers a single tool: mask_turkish_pii, which masks personal data in Turkish text. You run the text through this tool before sending it to a language model; the masked version is what reaches the model. No key is required. The transport layer is Streamable HTTP; there are no sessions and no SSE, each request stands on its own.

The tool masks all 26 categories; you can find the full list on the Categories page. In the output, values are replaced with numbered placeholders such as <KİŞİ_1>; the format details are on the Concepts page.

Adding to Claude Code

One command is enough:

claude mcp add --transport http siper-mask https://siper.gurubase.ai/api/mcp

After setup, Claude Code can call the mask_turkish_pii tool on its own when needed; you can also trigger it manually by saying “mask this text”.

Other clients

Setup links and configuration examples for Cursor, VS Code, and similar clients are on the info page that appears when you open the endpoint address in a browser: siper.gurubase.ai/api/mcp. A discovery card is also live so clients can recognize the server on their own: server-card.json.

Limits

This server is not the product itself but a demo service for evaluation; it introduces itself to clients under the name siper-mask-demo.

The input (text) is at most 2,000 characters.
The number of requests is limited (rate limit).
It runs without a key; for that reason it is not a production surface.

For production integration, use the product endpoints, not the MCP demo: the Gateway’s /v1 endpoint for automatic masking in a chat flow (Gateway integration), and the /mask API to use masking as a standalone step (Mask-only usage).

If you tried the demo and liked the output, the next stop is Getting started: you get your virtual key and send your first masked request from your own code.