Something that keeps cropping up more and more with each new integration, MCP and LLM tool, is trying to direct your LLM (and users) to the “correct” fields on any given table.
Ideally, you don’t hold on to old fields that aren’t used, but that’s not always super realistic.
I would like a way to add meta-data properties to fields (thinking this would be tenant level not object by object).
Here’s a general example:
- I’m setting up a Claude skill for my org and I want it to reference the correct fields. But if I “deprecate” a field, I then need to make sure my skill is updated and reading the right field.
- If I can do something even as simple as an “Active/Inactive” boolean at the field level, I could potentially flag those more easily for an LLM to ignore.
- As admin capabilities improve, I could potentially run internal meta-data updates to make bulk changes to fields, audit what is/isn’t used on layouts and see what maybe needs to be active/inactive.
I might want my own properties beyond a single boolean, but this would this could help admins manage data better at our orgs with a layer of platform level control beyond having to delete fields to get people and tools to “ignore” them.