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

# What is MCP

The [Model Context Protocol (MCP)](https://modelcontextprotocol.io/) is an open standard, maintained by Anthropic, that lets a host LLM discover and invoke **tools** exposed by third-party servers — with the end-user's own credentials. It is K-AI's primary surface for the **Activate** discipline: making governed, cleaned documents reachable by AI agents under the user's permissions.

## Why K-AI exposes MCP

K-AI is a Document Knowledge Platform. Govern and Clean produce a high-quality Neural Semantic Graph; Activate makes that graph reachable. We want a Consumer's LLM client — Claude Desktop, Cursor, Mistral Le Chat, or any MCP-compatible agent — to answer questions and run workflows on top of K-AI knowledge, without us dictating an orchestration strategy or keeping chat state server-side.

Instead of a single imposed `/search` endpoint, K-AI exposes **primitives** — list instances, search semantic nodes, fetch documents, drive audits — as MCP tools. The host LLM does the orchestration. The user's OAuth 2.1 credentials carry group-based permissions, instance visibility and per-instance instructions into every tool call.

This is why `/search` and `/conversation` were removed from the Instance API. See [Removed endpoints](/knowledge-ai/sources-and-ingestion/instance-api/removed-endpoints.md) for the full story.

## Retrieval MCP vs Audit MCP

| Aspect     | Retrieval MCP                             | Audit MCP                                           |
| ---------- | ----------------------------------------- | --------------------------------------------------- |
| URL        | `https://api-retrieval.kai-studio.ai/mcp` | `https://api-audit.kai-studio.ai/mcp`               |
| Tool count | 5                                         | 21 (curated subset of the Audit API)                |
| Shape      | Read-oriented (search, fetch)             | Workflow-oriented (conflict-first, state mutations) |

Both servers share the same OAuth 2.1 Authorization Server (`auth-api.kai-studio.ai`) — a token obtained from one is accepted on the other.

## Next

* [Connect a client](/knowledge-ai/k-ai-mcp/connect-a-client.md) — install in Claude Desktop, Cursor, Le Chat.
* [Schema stability policy](/knowledge-ai/k-ai-mcp/schema-stability.md) — what we commit to keep stable.
* Protocol details: [modelcontextprotocol.io](https://modelcontextprotocol.io/).