Documentation

Model Context Protocol Configuration

In order to use your query pipeline (see Query Pipelines ) for agents and AI, then you can use the RheinInsights Retrieval Suite as a Model Context Protocol (MCP) server.

Please note that as of now, our RheinInsights Retrieval Suite only exposes public search via MCP.

Please note that the RheinInsights Retrieval Suite supports streamable http as transport type. SSE as deprecated transport type is not supported.

Configuration

As a prerequisite, you need to have configured at least one public search query pipeline. Please refer to Query Pipelines .

Then please open Administration > Search UX > Bot Integrations

Click on “Add” to add a new bot integration.

MCP Configuration Dialog in the RheinInsights Retrieval Suite's admin interface

  1. Bot type: please choose Model Context Protocol (public search)

  2. Then choose the according query pipeline

  3. Finally, define the endpoint URL where the RheinInsights Retrieval Suite and the MCP endpoint will run. The MCP endpoint is always relative to the Suite’s base URL at /api/v1/mcp, so you need to add <baseurl>/api/v1/mcp here

After clicking on validate and save the configuration is stored.

Links Section

In the links section you will find two links.

  1. A yaml file to integrate the MCP server into PowerAutomate and Copilot.

  2. A json file which you can integrate into your local Claude instance.

Testing

After saving the configuration, you can test the enpdoint. For instance with the model context protocol inspector by running

npx @modelcontextprotocol/inspector (suppose you have nodejs installed)

you will be able to connect via <baseurl>/api/v1/mcp with transport type streamable http.

Please note that you need to have a valid https certificate for testing. In this case you either need to replace the https certificates or work with a proxy server or a data gateway like ngrok or zrok.