Designing an MCP Server for Technical Documentation
- Presentation
- Artificial Intelligence (AI) in Technical Communication
- 10. November
- 04:45 PM (CET) - 05:30 PM (CET)
- Plenum 1
-
Martin Blumbach
- Ericsson
Contents
Technical documentation is a valuable enterprise data source. It is nowadays usually provided to AI agents using a Model Context Protocol (MCP) server. In an enterprise context not all use cases can be known, but the goals of AI agents when interacting with technical documentation may not differ so much from the goals of human beings.
We'll present the status of an ongoing project at Ericsson, show how we approach the design of an MCP server for technical documentation, and explain our design decisions for MCP tools, resources, and prompts. We take into account that technical documentation consists not only of documents, but also of structured data (for example configuration parameters, counters, part lists). We'll share what we have learned so far in the hope of triggering a conversation on best practices for designing such MCP servers.
Beside the MCP server itself, we'll look at AI agent skills necessary to use it, and a high level AI knowledge and content strategy to support it.
Takeaways
- AI agents and human beings have similar information needs, but the means to access technical documentation differ
- How to use the Model Context Protocol (MCP) for technical documentation
Prior knowledge
- Basic knowledge of AI agents and the Model Context Protocol (MCP)
- Good understanding of how technical documentation is published