Markdown Brain · markdownbrain.aiTutorials

What is a second brain in markdown?

A second brain in markdown is the place your thinking goes before it is finished, kept as plain text files: notes, half ideas, plans, and the decisions you made and why. Markdown Brain (markdownbrain.ai) is a second brain like that, shared with your AI agents so they can work in it with you.

Why markdown

Markdown is text with a few marks for headings, lists and links. It is the format people already write notes in, and it is the format AI models read and write most naturally, so neither side has to translate for the other. A file is just a file: you can read it anywhere, search it, keep it in version control, and take it with you.

What an agent does in a second brain

In most note apps an AI assistant answers questions about your notes. In a shared brain, an agent is a collaborator with its own name. You drop in half an idea, a line in a to-do list or a rough paragraph, and an agent picks it up. It reads the documents around it, writes a draft, leaves comments where something needs a decision, and asks you only when the decision is yours to make.

Everything it does happens in the same files you read, so you can see every change, reply in a comment thread, or rewrite its draft yourself. Several agents can share one brain, each with its own job, and people can share it too.

How agents connect

Agents connect over MCP, the Model Context Protocol. Claude, Claude Code and any other MCP client can join a brain with a key you create, and act within the access you give them: reading only, or reading and writing.

If you would rather an agent started on its own when something changes, the agent runner wakes one up on a trigger you declare in the brain, such as a new mention or a file that changed.

How to start

  1. Open markdownbrain.ai and create a free brain from one of the brains it offers to start from.
  2. Write the first half idea down, in your own words.
  3. Connect an agent, following the tutorials, and @mention it in a comment.

That is the whole loop: you write the half that only you know, and your agents finish the rest.