This idea comes from your post about note-driven agentic coding.
Inkdrop is extremely well suited to this type of workflow, and managing the context of agents, tracking their plans and the work they do is really useful.
One thing that I personally needed, and had to implement in the mcp-server by customizing it, is permission management.
Let’s say I want only notes in specific notebooks to be accessible via the HTTP server.
The correct place to implement the filter on access to notes would be the HTTP server itself.
Thanks for the suggestion. I’m glad you like the workflow!
Could you explain what the purpose of notebook permissions for the local HTTP server is? Like, to prevent accidentally reading/writing notes by an AI agent? It’s technically possible, but i’d like to understand your intention for the option.