MCP in Practice: How LLMs Access Your Data in Context.
Friday, 13 November 2026 | 11:00 CET
Hosted by Niklas Endnich, Technical Success Manager
The Lobster Data Platform includes an MCP server, built directly into the API Manager. You can use it to offer data and processes from the platform as tools that AI assistants such as Claude or ChatGPT can call on their own.
With the MCP server, the integration team decides exactly which data and actions an AI agent may use. It runs on the Lobster logic and authentication they already have. Any MCP-capable client can connect, so business users get their answers in the AI assistant and the systems behind it stay as they are.
In this webinar Niklas will walk you through building an MCP server step by step, starting with a simple read tool and ending with tools that create and update data.
What you’ll take away:
- How MCP works, and how an MCP server differs from a REST API in the API Manager: you define tools instead of endpoints, and each tool’s name and description decide when the AI calls it
- How to build read and write tools (search, filter, create, update) with the Event Handling actions you already know, and how to return answers and error messages the AI can understand
- How to secure the server with an OAuth2 partner channel and a dedicated technical user, and why anonymous access belongs only on test systems
- How to connect your server to an MCP client such as Lobster GPT or Claude and test it
For whom this is suitable
Integration developers and solution architects who build REST APIs or Event Handling in the Lobster Data Platform and want to make their data available to AI agents. You should already know the API Manager and the common Event Handling actions (Search, Set Value, If-Then-Else, Create Instance). You don’t need any prior experience with AI or MCP.
Can’t join live? Register anyway – we’ll send you the recording.
