Skip to content
tuckit

Try it in ten minutes

By the end you will have a board, an agent attached to it, and a card the agent filled in itself. About ten minutes.

What you need

Section titled “What you need”

A browser. Steps 1 to 3 all happen on the web. Stop after step 3 and you still have a working board.

A coding agent on your machine. This page uses Claude Code because it takes the fewest steps. Codex CLI and Antigravity work too, with a different install command. Connect your agent covers each one.

Claude Code runs in a terminal, so from step 4 onward you are typing in a terminal rather than clicking in a browser.

  1. Sign up

    Section titled “Sign up”

    Go to app.tuckit.dev/register/ and sign up with an email and password, or with Google or GitHub.

    You’ll be asked for two names. One is your organisation’s name, the other is a short lowercase one for the address. Type acme and your board lives at /acme/board/, and every card gets a number like ACME-1, ACME-2. Paste one of those numbers at an AI and there is no ambiguity about which card you mean.

    You need an organisation even working alone. Your project name or your company name is fine.

  2. Make one area

    Section titled “Make one area”

    An area (Area) is a broad heading: frontend, backend, docs. It changes rarely and does nothing except hold cards.

    Make one. You can add more later, and moving cards between areas is free.

  3. Write down one thing to do

    Section titled “Write down one thing to do”

    Create a card (Slice). Give it a title and leave everything else alone.

    Leave the “what we’re building” box (Spec) empty on purpose. Empty is how the board says this hasn’t been worked out yet, and it’s the first thing your agent notices in the next step.

  4. Attach your agent

    Section titled “Attach your agent”

    Open a terminal, go to a project folder, and start Claude Code by typing claude. At its prompt, run these two lines:

    /plugin marketplace add tuck-it/tuckit-plugins
    /plugin install tuckit@tuckit-plugins

    The connection comes with the plugin, so there is nothing else to set up. If it asks for an address, it’s already filled in; press Enter.

    The first time your agent uses tuckit, a browser window opens and asks whether to allow it. Press Allow. There is no long string to copy from one place and paste into another.

    Installing this also gives you three things: a note at the start of each session telling the agent to read the board first, a nudge at the end to write things back, and a set of named ways of working the agent loads when it needs them. What you install walks through them.

  5. Ask it to read

    Section titled “Ask it to read”

    Just ask:

    What's the state of this project?

    If it answers with your organisation, the area you made, and how many unsorted items are waiting, it’s connected. If it answers by reading files or commit history, it isn’t. Check /plugin list.

  6. Ask it to write

    Section titled “Ask it to write”

    Now point it at the card from step 3:

    Take the card I just made, work out what to build, write it into the card,
    and write down what we'd have to see to call it done.

    Refresh the browser. The “what we’re building” box is filled in, there’s a sentence in “done when” underneath, and the card has moved two columns to the right — past Done when? and into In progress.

    Nobody moved that card. It moved because of what’s now inside it.

That’s the setup finished

Section titled “That’s the setup finished”

Home shows a connect an agent panel until an agent has actually written to the board. That is the only thing it waits for, and it is the one thing the plugin cannot do on your behalf — something has to connect before it can write anything. Once step 6 lands, the panel takes itself away.

Where to go

Section titled “Where to go”