New · Model Context Protocol

The Sonat MCP Server

One endpoint — mcp.sonat.com — that turns your knowledge base into a live tool for any AI assistant. Connect Claude, ChatGPT, and Cursor so they can search, answer, summarize, and even publish against your own documentation.

Get Started for Free Talk to Sales
claude_desktop_config.json 
{
  "mcpServers": {
    "sonat": {
      "url": "https://mcp.sonat.com/mcp",
      "transport": "http"
    }
  }
}
tools/list search · ask · summarize · rewrite · create · publish
Why it matters

Your documentation, now an AI-native tool

The Model Context Protocol is the open standard that lets AI assistants talk to external systems. The Sonat MCP Server implements it for your knowledge base — so instead of copy-pasting docs into a chat, your team's AI reads the source of truth directly, and answers stay current as your content evolves.

Connect the AI clients your team already uses

One standards-based endpoint. No custom SDK, no per-client rebuild.

Claude

Add a custom connector and authorize — the reference MCP client.

ChatGPT

Developer-mode connector; search + fetch map cleanly to your KB.

Cursor

Bring your documentation into the editor your team already uses.

Any MCP client

Streamable HTTP + JSON-RPC 2.0 — the open standard, not a lock-in.

What your AI can do

A curated, vetted set of tools — not an auto-generated firehose. Reads and AI skills are on by default; writes are unlocked only for tokens with a write role.

🔍

Search & read

Find and pull the exact content your AI needs to answer accurately.

  • search_knowledge_base Full-text search across the account's published content.
  • ask_question RAG answers grounded in your documents, with citations.
  • get_document Fetch a document by id.
  • get_topic_content Read the content of a single topic.
  • get_topic_tree Walk the published topic hierarchy.
✨

AI skills

Sonat's own content intelligence, callable as tools.

  • summarize Condense a topic or document into a brief.
  • rewrite Rewrite content for tone, clarity, or length.
  • generate_content Draft new content from a prompt and your context.
✍️

Create & publish

gated

Draft, edit, and ship documentation — only with a write-scoped token.

  • create_document Create a new document (slug derived from the title).
  • update_document / update_topic Fetch-merge edits — only the fields you send change.
  • publish_topic Publish the latest draft of a topic.

Live in three steps

1

Mint a token

Generate a Sonat API token scoped to the account (and, optionally, the specific documents) you want to expose.

2

Add the connector

Point your AI client at mcp.sonat.com/mcp and authorize. The tenant is resolved from your token — no server config to run.

3

Ask & act

Your AI now searches, answers from, and (if permitted) writes back to your knowledge base — grounded in real content.

Built for trust

Enterprise-grade by design

Exposing your knowledge base to AI shouldn't mean loosening control. Every request is bound to a single tenant and re-checked downstream — an AI can only ever touch what its token is allowed to touch.

🛡️

Strict tenant isolation

Account and version ids are injected server-side from your token — never trusted from the client. Cross-account access is rejected.

🔐

OAuth 2.1 + PKCE

Native connectors authorize over the standard the MCP spec mandates, on top of Sonat's existing auth.

🎛️

Gated, curated tools

Writes require a write role and are enabled deliberately. Destructive actions are flagged and off by default.

📋

Audited & routed

Calls flow through the Sonat gateway, reusing its logging, routing, and IP controls.

Bring your documentation to every AI

Keep authoring in Sonat. Let your team's AI assistants read and act on it, safely, through one standards-based endpoint.

Get Started for Free Request a Demo

Ready to create your first manual with Sonat?

Free Sign-up Request Demo