> For the complete documentation index, see [llms.txt](https://zcloak-hub.gitbook.io/zcloak-ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://zcloak-hub.gitbook.io/zcloak-ai/products/mcp-server.md).

# MCP Server

### What zCloak AI MCP Server Is

The **zCloak AI MCP Server** bridges conversational AI platforms to the ATP ecosystem.

It allows AI systems to:

* Verify identities
* Publish claims
* Sign commitments
* Query the Data Plane

All without embedding protocol logic directly into the model.

***

### Why MCP matters

Large language models:

* Speak natural language
* Do not enforce schemas
* Cannot safely sign cryptographic data

MCP provides a **trusted middleware layer** that:

* Validates intent
* Enforces protocol structure
* Handles signing securely

***

### Core capabilities

* **Schema validation**: ensures events conform to ATP spec
* **Signing operations**: interfaces with key management for secure signing
* **Event translation**: converts conversational commitments into structured ATP events
* **Query interface**: search and verify identities/claims on the Data Plane
* **Interoperability**: follows MCP conventions for platform compatibility

***

### Typical usage flow

1. User negotiates with an AI Avatar in conversation
2. Avatar drafts an ATP event through MCP
3. Event is validated, signed, and published
4. Counterparty verifies and (if needed) signs via embedded web URL