FOR AI AGENTS · MCP

Let your assistant
write to your canvas.

A small bridge that connects Claude Code, Claude Desktop, and any MCP client to your boards. Your assistant can create notes, search what you've written, and see your canvases. You point it at a board and it does the writing.

Connect it open the canvas
WHAT YOU CAN DO

Just ask. Notes appear where you'll find them.

YOU

Summarize this thread and put it on my Research board.

YOU

Compare these three libraries in a table on Research.

YOU

What have I already written about pricing?

YOU

Save this SQL snippet to Snippets so I don't lose it.

THE TOOLKIT

A small toolkit your agent can reach for.

It picks whichever it needs. Everything it writes is plain markdown. It renders exactly like a note you typed yourself and is instantly searchable.

list_boards

Find your canvases

Lists every board by name and id, so the agent knows where a note can go before it writes one.

create_board · rename_board · delete_board

Manage boards

Spin up a fresh canvas, rename one, or delete it. Deleted boards and their notes stay restorable in the app for 30 days.

create_note

Write a note to a board

Drops a markdown note onto the board you name, with headings, task checkboxes, bold, code, links, and images. The position is optional.

list_notes · update_note · delete_note

Edit the canvas

Read a board's notes with their ids, rewrite a note's text, move it, or delete it.

create_task · update_task

Track its work as task cards

A live job card goes queued → running → done on your canvas. When it finishes, the result settles in as a plain note.

create_object · set_object_state

Fill live tables and embeds

Drops a table you can keep editing by hand, or a live embed — YouTube, Spotify, Figma — and updates it as it works.

search_notes

Search what you've written

Full-text search across all your notes, returning highlighted snippets so the agent can find and quote what's already there.

IN PRACTICE

Where it earns its keep.

Researching in Claude
Each finding drops onto a Research board as it goes.
Planning a feature
The plan lands as a task list you can tick off on the canvas.
Debugging
The working fix and its context get saved to Snippets for next time.
Reviewing your notes
The agent searches everything you've written and cites the right note.
Clearing your head
A brain-dump gets captured to Inbox mid-conversation.
Comparing options
A live table lands on the board, and you can keep editing it by hand.
Running a long job
A task card tracks it queued → running → done, then settles into a note.
SETUP · ABOUT FIVE MINUTES

Connect it once.

01

Mint a personal token

In the app, open the command palette (⌘K / Ctrl-K), run API tokens, name it, and hit Create. Copy the jnt_… secret. It is shown once. Revoke it anytime to instantly cut off access.

02

Nothing to install

The server is published as @codellyson/justanotetaker-mcp, so your agent runs it straight from npm via npx. There is nothing to check out or build.

03

Register it with your agent

Add it to your MCP config: .mcp.json for Claude Code, or the Claude Desktop config. Restart, and the tools show up.

{
  "mcpServers": {
    "justanotetaker": {
      "command": "npx",
      "args": ["-y", "@codellyson/justanotetaker-mcp"],
      "env": {
        "JUSTNOTE_TOKEN": "jnt_…",
        "JUSTNOTE_API_URL": "https://api.justanotetaker.kreativekorna.com"
      }
    }
  }
}

JUSTNOTE_API_URL is optional. Set it to your own deployment if you self-host the API.

YOURS, AND ONLY YOURS

How the token keeps it safe.

Runs locally

A stdio server that the agent launches as a subprocess. Nothing listens on a network port.

Scoped to you

The token sees only your account's notes and boards. It can't touch your account or password.

Shown once, revoke anytime

Stored server-side as a hash. Revoke it and whatever was using it stops immediately.

HANDS-FREE NOTES
Let the assistant
do the filing.