Our small company works in the data space, and the shift towards agent-friendly data surfaces is the big thing right now.We decided to get stuck in by learning how to build MCPs and semantic layers, and then testing them with our clients. To be honest, we wouldn’t say this is too challenging for a smart data engineer or analyst with a good agent client.However, adding charts (as an MCP app) and writing the front-end code was a pain. Trying to make these agent-friendly was also a pain. After lots of trial and error, we’ve put something fairly effective and wanted to share it with the wider data community.If you’re building MCPs on top of any sort of data, then give it a try. We’re always open to feedback