Knowledge base as an MCP server
The Model Context Protocol (MCP) is an open standard that lets AI clients use tools from external systems. vragen.ai makes your knowledge base available as an MCP server, so a client such as Claude can search your knowledge base and read documents on its own. This lets you use your own, reliable content as a source within an AI assistant, without building a custom integration for it.
Endpoint
The MCP server runs on your own subdomain and uses HTTP as transport:
https://{klant}.vragen.ai/mcp/knowledge-base
Authentication
The server works with an API token that you create and manage in your vragen.ai environment. You pass it as a Bearer token, just like with the REST API.
Available tools
The server offers an AI client the following tools:
- Search the knowledge base: a semantic search of your knowledge base, so the client can retrieve relevant passages.
- Read a document: read (parts of) a found document to substantiate the full answer.
- Retrieve filters and facets (depending on your configuration): retrieve the available filters and their values, so the client can search more precisely.
Connecting
In your MCP client, add a server with the endpoint URL above and your API token as a Bearer token. The client then discovers on its own which tools are available. Before going live, you can test the connection with an MCP client or an MCP inspector.
Support
Stuck? Email our support team at service@vragen.ai.