Hi @Bill_French I’m very interested in how you gave the request to the Coda MCP Server. Is this different than using the Coda AI assistant because I tried that and didn’t get anything like the results you posted.
Also, I don’t fully understand how to even ask the question to AI other than the post that I made here.
From a user perspective, this also brings up a couple of interactive issues:
-
Why not have the Coda MCP server reply to each and every post in Ask the Community? This wouldn’t stop other users, especially if the reply is labeled as an AI agent and would help with making the AI agent better.
-
Why not build the AI support agent directly into Coda so that it can look at a doc and with a little help from the user, determine what the user is trying to do. The Coda AI agent could even go through a doc, looking for any formulas that aren’t working and offer to help with each one.
The fact that the MCP gave an appropriate response it proof of the underlying intelligence but without making it easy for users it hamstrings the growth of Coda.
Please also ensure that the MCP is programmed morally as well, wouldn’t want it to be like the Tron MCP.