Very cool that we have a MCP server, we have been wanting this for a while because some people have asked about it.

But I am not sure why it is needed for the agent project. Is this an optimization, or is it strictly necessary?

I am very concerned about this new API which publicly exposes the subscribe/unsubscribe features. It is kind of a red flag if anybody can subscribe or unsubscribe from any content using a public api.

https://github.com/seed-hypermedia/seed/compare/knowledge-agent-server-setup#diff-b6d30057efa25c2fb6736ed8531a61e21cbca45732c079fd9f2f03ed493f1b9aR64-R66

Instead of doing this, can we just use the gRPC skill, which is designed to give an agent private/secure control over a daemon.