MCP Support for Collected Notes: AI-Powered Writing Workflows

I just finished building an MCP (Model Context Protocol) server for the upcoming version of Collected Notes 3, and it unlocks some genuinely exciting possibilities for AI-assisted writing workflows.

What We Built

The MCP server exposes the full Collected Notes API to AI assistants like Claude. This means any AI tool that supports MCP can now:

  • List and search notes across all your sites
  • Create new notes directly from conversations
  • Update existing content without leaving your AI workflow
  • Semantic search using AI embeddings to find notes by meaning, not just keywords

What You Can Build With This

1. Conversational Publishing

Write a blog post by talking to your AI assistant. Describe what you want to write about, iterate on the draft in conversation, then publish directly—all without opening a browser or editor.

2. Knowledge Base Assistant

Connect your notes to an AI that can search them semantically. Ask questions like "What did I write about authentication?" and get answers synthesized from your own content.

3. Cross-Platform Content Sync

Use the MCP server as a bridge. AI assistants can read content from one source (a GitHub repo, a Notion page, a conversation) and publish it to Collected Notes automatically.

4. Automated Content Updates

Build workflows that keep content fresh:

  • Update a "Now" page based on your calendar
  • Refresh a "Reading List" from your bookmarks
  • Sync release notes from your changelog

5. Research-to-Publication Pipeline

While researching a topic, have your AI take notes. When you're ready, ask it to synthesize those notes into a published article—with citations back to your sources.

6. Multi-Site Management

Manage multiple Collected Notes sites from a single interface. The AI can help you decide which site a piece of content belongs on, cross-post with modifications, or maintain consistent voice across properties.

Why MCP Matters

The Model Context Protocol is becoming the standard way for AI assistants to interact with external services. By building an MCP server for Collected Notes, we've made it a first-class citizen in AI-powered workflows.

This isn't just about convenience. It's about removing friction between thinking and publishing. The less time you spend context-switching between tools, the more you can focus on the actual writing.

The Technical Bits

The server implements these MCP tools:

  • list_sites / get_site - Site management
  • list_notes / get_note - Note retrieval with pagination
  • search_notes - Full-text search
  • semantic_search - AI-powered similarity search
  • create_note / update_note / delete_note - Content management

Each tool is designed to be composable. An AI can chain them together: search for related content, read it for context, then create a new note that builds on existing ideas.

What's Next

This is just the beginning. With MCP as the foundation, we can add:

  • Scheduled publishing - Draft now, publish later
  • Content analytics - Ask "which posts performed best?"
  • Collaboration features - AI-mediated editing workflows
  • Export pipelines - Publish to multiple platforms simultaneously

The future of writing tools isn't about better editors. It's about intelligent assistants that understand your content and help you create more of it.

This post was created using the very MCP server it describes. Meta, I know.

Powered by Collected Notes