Post-itSign in

Post-it

Post-it is a place to write things down, where who can read what is the product rather than a setting on the side of it. You write Markdown in spaces, organise it in folders, and share a page or a whole folder with a person, with a team, or with anyone at all. A page you have not been given is not merely hidden from you: it is indistinguishable from a page that does not exist.

The other half is that you can connect it to Claude, so the things you have written are available in conversation without copying them anywhere. That is what the rest of this page is about.

Connecting Post-it to Claude

1. Make a token

Sign in, then go to Your spaces and follow Connect Post-it to Claude, or go straight to /settings/mcp. Give the token a name you will recognise in six months, like the machine it is going on.

The token is shown once and never again. Only a hash of it is stored, so nobody, including us, can recover it. If you lose it, revoke it and make another; that takes ten seconds and is the right instinct. Revoking takes effect on the very next request.

By default a token reaches everything you can reach. You can pin one to a single space at creation, which is worth doing for a machine you trust less than your own.

2. Point Claude at it

Your endpoint is:

MCP endpoint
https://post.maqsoodlabs.com/api/mcp

Claude Code. Run this in a terminal, or put the JSON in .mcp.json at the root of a project.

Claude Code, command line
claude mcp add --transport http postit https://post.maqsoodlabs.com/api/mcp \
  --header "Authorization: Bearer post_your_token_here"
Claude Code, .mcp.json
{
  "mcpServers": {
    "postit": {
      "type": "http",
      "url": "https://post.maqsoodlabs.com/api/mcp",
      "headers": {
        "Authorization": "Bearer post_your_token_here"
      }
    }
  }
}

The Claude apps, desktop and web. Settings, then Connectors, then Add custom connector. That dialog takes a URL and nothing else, so for these the token goes in the URL itself: https://post.maqsoodlabs.com/api/mcp/your-token, which the token screen gives you ready to paste.

That form is deliberately the weaker one. A token in a URL is in every HTTP log that records the path, in whatever the client stores for the connection, and anywhere the URL is pasted; a token in a header is in none of those. Use it only where a header is not on offer, and prefer a token pinned to a single space so that a leak costs one space rather than an account.

The API. Three things are needed and it is easy to give two: the beta header, the server in mcp_servers, and a matching mcp_toolset entry in tools. Leaving the toolset out does not quietly ignore the server, it makes the request invalid.

Anthropic API
curl https://api.anthropic.com/v1/messages \
  -H "content-type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "anthropic-beta: mcp-client-2025-11-20" \
  -d '{
  "model": "claude-opus-5",
  "max_tokens": 2048,
  "messages": [
    {
      "role": "user",
      "content": "What is on my todo list?"
    }
  ],
  "mcp_servers": [
    {
      "type": "url",
      "url": "https://post.maqsoodlabs.com/api/mcp",
      "name": "postit",
      "authorization_token": "post_your_token_here"
    }
  ],
  "tools": [
    {
      "type": "mcp_toolset",
      "mcp_server_name": "postit"
    }
  ]
}'

What a connected Claude can and cannot reach

A token acts as you. Every request it makes is answered by the same rules that answer your requests in the browser, so a connected Claude sees exactly what you see and never more. Share a folder with yourself tomorrow and it appears; have it revoked and it disappears, with no cache to go stale in between.

It can:

  • list_spaces, list_tree to see what is in one, search within it, and read_page by path or id
  • list_skills and get_skill, so it can find and follow the skills you have written
  • create_folder and create_page, so it can build structure and not just a flat list, and update_page guarded by a version number, so it cannot overwrite an edit you made while it was thinking
  • list_backlinks, to see what points at a page

It cannot:

  • delete or move anything
  • change who can see anything
  • reach a page nobody has given you, whatever it is asked to do

That list is short on purpose. A token is a long-lived credential sitting on the internet, and one that can only add is a far smaller problem to have leaked than one that can remove. The dangerous things are still available, deliberately, in the browser.

Articles and skills

Every page in Post-it is an article or a skill. An article is prose: notes, a decision, a write-up. A skill is a Markdown file written as instructions for Claude to follow, in the same shape Claude uses elsewhere, which means Post-it doubles as a place to keep them.

The distinction earns its keep the moment you ask for your skills:

  • list_skills returns your skills and nothing else, so "what can you do for me here" does not also return every meeting note.
  • get_skill returns the instructions with the metadata stripped, ready to follow.

A skill worth having:

  • Has a name and a description in its frontmatter. The description is what a client reads to decide whether the skill is relevant, so it should say when to use it, not what it is called.
  • Says what to do, not what the topic is. "Read the runbook before starting, every time" beats "runbooks are important".
  • Includes a worked example. A skill nobody can picture using does not get used.
  • Names the failure it exists to prevent. That is usually the reason somebody wrote it, and it is the part that makes the instruction stick.

Post-it will not stop you saving a skill with no description. It will say so and save it anyway: losing what you wrote over a formatting detail is a much worse outcome than an incomplete skill.

Six skills to start with

Copy any of these into a space, mark it as a skill, and it is available to your Claude immediately. They are meant to be edited: the house style one in particular is worth nothing until it describes how you actually write.

Chat Context

File what a conversation decided into Post-it, and pick it back up in a later session.

Chat Context

Todo List

Keep todos as a checklist in a page. Claude adds items, reads what is outstanding, and ticks things off.

Todo List

Decision Log

Record a decision with its alternatives and what would make you revisit it, then answer 'why do we do it this way' from the record.

Decision Log

House Style

Teach Claude how you write, so drafts come back sounding like you rather than like a press release.

House Style

Onboarding a Colleague

Share a folder with somebody and their own Claude picks up the project's context, through their own token.

Onboarding a Colleague

Runbooks

Operational procedures written once and followed consistently, read live rather than from a stale copy.

Runbooks

Two questions people ask

If I publish a page, does Claude see the comments on it? No. Publishing a page puts the document on the internet; the conversation about it stays behind a login. Comments tend to be candid in a way the document is not, and one toggle should not put them in public.

Can somebody work out that a page exists by searching for a word in it? No. Search runs under your permissions, so a word that appears only in a page you cannot read returns nothing at all, rather than a result with the content hidden. The same is true of links: a link to a page you may not read renders identically to a link to a page that does not exist.