From zero to a connected team in ten minutes.

Each chapter plays a short run-through on Laminary's own screens, with the same steps written out underneath. Names and numbers in the examples are a sample team.

Chapter 1

Create a team and a project

A team is the people who share one ledger. A project is where a piece of work comes together: a game, an app, a client site. A repo is optional.

  1. Sign in at the control room with email or Google.
  2. Name your team and create it. You're its owner.
  3. Press + next to Projects, name the project and choose who's on it: everyone, some people, or just you for now.
  4. Send teammates the invite link from the Team page.
Sample team
Chapter 2

Connect your AI

Everyone connects their own tools, once. Each connection has its own key you can revoke without touching the others. There's one address for every tool:

https://app.laminary.dev/mcp

  1. In the control room open You and your tools, then Add a tool, then Claude or ChatGPT, and copy the address.
  2. In Claude: Settings, Connectors, Add custom connector, and paste the address. In ChatGPT: add a custom connector under Settings, Connectors (it may ask you to turn on developer mode first).
  3. Press Connect. Laminary's page opens; pick the team and press Allow.
  4. Back in the chat, ask anything. The connector shows Laminary's name and icon.
Sample team
Chapter 3

One project per chat, set once

With several projects, a chat needs to know which one it's in. Put one line in the project instructions and every chat in it binds by itself. A chat without the line asks you once.

  1. Open Project settings and copy the line under "Use this project from a chat".
  2. In Claude, open your Claude Project, then Instructions, and paste it. In ChatGPT, use the project's instructions.
  3. New chats in that project land in the right Laminary project with no question.
  4. Claude Code works it out from the repo. To pin a repo, run laminary link store-launch.
Sample team
Chapter 4

While you work: skip, or stop before overlapping

Nothing new to learn. Before each task your AI checks the ledger and tells you what it found, in one sentence, in the chat you already have open.

  1. You ask your AI for something as usual.
  2. If a teammate already finished it, your AI takes their result instead of redoing it.Cursor, after checking LaminaryLucas built the chip system an hour ago. Using his files instead of making them again.
  3. If a teammate is on it right now, your AI stops and asks.ChatGPT, after checking LaminaryIdan is on the 3D intro right now. Take something else, or wait for his result?
  4. If it's free, your AI claims it so nobody else starts it.
Sample team
Chapter 5

Handoffs: automatic, to a person, or on request

Work doesn't stop when an AI does. A handoff is what's done, what's left and the next step, so a teammate's AI continues instead of starting over.

  1. Automatic. An AI runs out of tokens or the chat goes quiet. Its last checkpoint becomes a handoff and the team gets a notification.
  2. In Needs you, press Continue with my AI, or tell your AI "continue Lucas's handoff". It picks up from the next step.
  3. To a person. Tell your AI "hand this to Maya". She gets a notification and her AI offers it at her next turn.
  4. On request. Type "ask Lucas to send me the intro" in the control room's ask bar, or tell your AI. Lucas's AI brings it up and hands it back to you.
Sample team
Chapter 6

Tasks and files, pulled straight into a chat

Plan work in the control room and attach what it needs: drafts, images, a Figma link. Any connected AI can pull the task with its files.

  1. Open Tasks, press New task, give it a title and a description of what done looks like, and assign it.
  2. Attach files or links. The assignee gets a notification.
  3. In any chat: "what are my tasks?", then "pull the store page task". Your AI gets the description, small text files inline, and download links that work for an hour.
  4. "Take it" links the work to the task. The task moves to Doing, and to Done when the work is complete.
Sample team
Chapter 7

What's shared, and what stays on your machine

No code is shared, ever. Each session has a sharing level, and you can change it any time from the control room or the CLI.

  1. Private: nothing leaves your machine. Skips and handoffs don't work for that session.
  2. Announce: the goal and status only, so teammates get overlap warnings.
  3. Summary (the default): plus files touched, outputs and the handoff. This is what makes reuse work.
  4. Full: plus the chat lines, for teams that want them.
Sample team
Chapter 8

Reading the control room

Keep it open in a tab. You'll get a notification when something needs you; the rest is there when you look.

  1. Needs you: handoffs for you, requests, overlaps, assigned tasks.
  2. Flow: one lane per teammate, left to right up to now. Green curves are reuse, dashed orange ones are continued handoffs.
  3. Ask bar: "is anyone doing the 3D intro?", or a command like "ask Lucas to send me the intro".
  4. Saved this week: tokens not spent twice, counted from the original sessions.
Sample team

That's all there is. Your AI does the rest.