MCP was not clicking for me, so the best way to learn was to build a thing with it. I have a plethora of notes everywhere and now I have a machine that drops rough ideas into my publishing backlog.
MCP is basicly an api for AI that tells agents how it's supposed to be used.
Kalle Tolonen
16.9.2026
Assistants can fill out ideas for backlog. Humans should still do the writing that isn't, for example, tech specs.
I need tech notes that are accessible for the tools that do the implementation. Publishing is a bonus that comes with minimal extra effort. I can go for a walk and talk with an AI and then ask it to post the summarized convo to my ideas.
Two routes to idea backlog:
ktcom-ideas) → backlogThe MCP-endpoint is auth protected. Tools under the hood:
create_article_idea / update_article_idea / list_article_ideas — backlog capture (title, notes→lead, optional text→body, tags, priority)promote_article_idea — turn an idea into a dashboard draft (idempotent if already promoted)list_draft_articles / get_draft_article / update_draft_article — read and patch drafts and scheduled posts (status draft | scheduled)get_article_authoring_guidelines — field limits, promote mapping, Markdown conventionsRough idea → create_article_idea (or update) → optional promote_article_idea → human writes and edits in the dashboard.
Real examples already sitting as published items or drafts on the site:
More moving parts (assistants, MCP, backlog, dashboard) vs a simple todo. Probably useful to only me, since this is an eccentric niche blog.
No published comments yet.
Your comment is held for review before it appears publicly.