First, I’m extremely excited to see the new MCP server features. They look excellent and exactly in the right direction for DT. Thank you so much for doing this.
Second, as currently implemented they’re not really granular enough in what they provide access to for me to get the most out of them.
I’m a lawyer. I use DT for my work (a database for public legal docs, a database per case) and a lot of personal admin. At the moment, I’m running it alongside Claude Code-produced set of tools that:
— maintain a parallel RAG database synced to the DT databases (I enable them one at a time; the semantic db stays in sync)
— provide MCP tools to query and expose DT data and metadata both via the RAG route and via JXA to DT search directly
— also offer various other MCP tools for custom legal research etc
I have set up ZDR use of some LLMs which I expose to DT’s AI interface for in-DT chat (not that I use that much, to be honest).
My homebaked DT MCP tools are gated: I can give access to chosen non-confidential databases to, say, Claude Desktop, and access to all dbs including confidential ones to ZDR models only.
I’d love to replace most of my own MCP tools with DT’s new ones, which I’m sure are better, faster and more extensive. But at the moment I can only exclude some databases entirely. What I’d love to do is have a way of gating some databases so that only agents with a particular token or other controllable permission can get to them, while making others available generally. Then I can use Claude Desktop for my life and general research and more compliant but also less polished things for the confidential stuff.
ps Also, I’m a little bit nervous of how easy it is to expose databases to Claude – it’s a single checkbox click away; and one has to exclude databases manually
pps obviously a built-in RAG tool some day would be amazing
