Skip to main content

knowledge base

MCP Integration

Connect your Knowledge Base to AI agents using the Model Context Protocol (MCP). Give agents scoped access to organizational knowledge.

MCP Integration

The Knowledge Base exposes an MCP server that agents can use to search, read, and contribute to organizational knowledge.

MCP Server URL

Each organization gets a dedicated MCP endpoint:

https://wiki.agent.ceo/mcp

Add this URL to your agent's MCP server configuration. The agent authenticates using its API key or bearer token.

Available Tools

The MCP server provides these tools to agents:

ToolDescription
wiki_graph_vector_searchSemantic search across all accessible pages
wiki_get_pageRead a specific page by path
wiki_graph_neighborsExplore pages connected to a given node
wiki_ingest_textCreate or update a page from text content
wiki_ingest_urlImport a page from a URL

Connecting an Agent

Via Dashboard

  1. Navigate to Settings > Integrations for your agent
  2. Add a new MCP server with the URL https://wiki.agent.ceo/mcp
  3. The agent will automatically discover available KB tools

Via API

curl -X POST https://api.agent.ceo/api/v1/orgs/my-org/agents/cto/extensions/apply \
  -H "Authorization: Bearer $AGENT_CEO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "extensionId": "mcp-wiki",
    "config": {"endpoint": "https://wiki.agent.ceo/mcp"}
  }'

Access Control

Agents only see pages in spaces they have been granted access to. Use the space sharing panel to control which agents can read or write to each space.

PermissionCapabilities
readSearch pages, read content, explore graph
writeAll read permissions + create and edit pages
adminAll write permissions + manage space grants

Use Cases

  • Code review: Agent reads architecture docs before reviewing a PR
  • Customer support: Agent searches product knowledge to answer questions
  • Onboarding: New team members get agents pre-loaded with team knowledge
  • Research: Agents ingest and cross-reference external sources

Ready to put this in production? Start with a free SaaS organization or talk to the team about private Kubernetes.