Skip to content

quickstart.md

Latest commit

 

History

History
151 lines (118 loc) · 7.08 KB

quickstart.md

File metadata and controls

151 lines (118 loc) · 7.08 KB
Raw

Quickstart

A walkthrough that takes a brand-new account from signup to a working chat with your first agent.

Signing in for the first time drops you into a full-screen guided setup. It walks you through the essentials — name your team, connect a computer, and create your agent — then hands off to your first chat, where the agent helps you finish getting set up. A progress rail on the left tracks where you are, and you can leave and resume later. This page mirrors that flow.

Self-hosting? Use your own deployment's URL wherever this guide says https://first-tree.ai. The rest is identical — the page gives you a short connect code, and FIRST_TREE_SERVER_URL can point the CLI at a custom deployment when you are not using the channel default.

Before you start

  • A GitHub account to sign in with.
  • A macOS or Linux computer where your agent will run. The shell installer bundles Node.js, so Node does not need to be installed separately.

1. Sign in and name your team

Open https://first-tree.ai and sign in with GitHub. A personal team is created for you automatically; the first step, Name your team, lets you confirm or rename it (you can rename it anytime later). Press Continue.

2. Connect a computer

Your agent runs on a real machine, so the next step links one to your team. Copy the command the page shows and run it in a terminal on that machine:

curl -fsSL https://download.first-tree.ai/releases/prod/install.sh | sh
~/.local/bin/first-tree login <connect-code>

This is the exact command for hosted production. Use the command from the page for staging or a self-hosted deployment: it selects the correct installer, binary, and server URL. The explicit ~/.local/bin path works before a shell reload adds that directory to PATH.

The first line streams the installer to sh; the second line signs the computer in. They are intentionally independent and do not provide shell-level transaction protection: when pasted together, an install-line failure does not automatically prevent the login line from running, and POSIX sh does not guarantee that curl | sh preserves a curl failure status. first-tree login:

  • Exchanges the short connect code against the CLI channel's default server URL. Use FIRST_TREE_SERVER_URL for a custom deployment. Connect URLs are not accepted; legacy JWT tokens still work during rollout.
  • Persists your member credentials and writes this machine's client.id.
  • On macOS / Linux, installs and starts a background daemon so the machine stays online across reboots. (See the Onboarding Guide for exactly what gets written and the --no-start flag.)

The page watches for the machine, then confirms a coding agent (such as Claude Code or Codex) is installed and signed in on it. Once it shows your computer connected with a ready coding agent, press Continue. If it reports that none is ready, install one (e.g. Claude Code) on that computer and sign in; it appears here automatically.

3. Create your agent

Your team agent runs on top of a coding agent already installed on the computer you just connected. Pick which one to use (Claude Code is selected by default when detected), give the agent a name (for example Buddy), and choose who can use it:

  • Shared with team — anyone on your team can talk to this agent.
  • Just me — only you can see and talk to it.

There is no "type" to choose — every agent you create here is a standard agent, and the visibility choice is the only difference. Click Create.

Setup then waits for the agent to come online on the computer you connected ("Bringing your agent online… usually a few seconds"). This readiness gate is what guarantees your first message reaches a live agent. The daemon on your machine picks up the new agent automatically — there is no CLI command to run by hand. If it doesn't come online within ~30 seconds, check that the computer is awake and its coding agent is signed in, then try again.

4. Start your first chat

The final step launches your first conversation and lands you in the Workspace with the chat open. Setup finishes here — agent-led, in the chat, not as more wizard steps. Your agent walks you through the key operations, one approved step at a time:

  • Point it at your code — share a local project folder path or a GitHub repo URL, and the agent reads your project and gets a first task done. (No GitHub setup is required up front; you can connect the First Tree GitHub App later, when a task needs it.)
  • Build your team's Context Tree — once it has shown real value from your code, your agent offers to build your team's shared memory: it proposes an initial structure and fills it in from your code, with you approving each change. (Offered to team admins; teammates who join later inherit the tree automatically.)

You reach a working agent first — none of this is a gate you clear up front. You can also connect code anytime from Settings.

The Workspace is a three-panel layout:

  • Left rail — the conversation list. New chats open inline; pick a target from the composer rather than a separate dialog.
  • Center — the selected conversation. Type a message and press Enter to send. Markdown is rendered live; agent-generated document references open inline in the right panel's preview mode.
  • Right — context for the selected chat: session state, agent text output, connected computer, coding agent, SDK version, recent notifications, and management links.

That's it — you're chatting with your first agent.

Invited to an existing team?

If you joined through an invite rather than creating the team, setup is shorter: a brief welcome, Connect a computer, Create your agent, then Start chat. There is no team-naming step — your team's admin has already set up the team, so your agent inherits whatever shared context it has (its connected code and Context Tree) automatically. Its first chat helps you get settled rather than building the tree from scratch.

Where to go next

  • Onboarding Guide — the same setup driven entirely from the CLI (login → agent create → daemon start), plus the SDK and troubleshooting reference.
  • CLI Reference — every namespace, every command, every env var.
  • Observability — how to wire metrics and logs into your stack.

Adding more later

The guided setup covers your first run. Afterward, manage everything from the console:

Need Where
Connect another computer Settings → Computers → + New Connection
Connect a code repo Settings → connect GitHub
Create another agent Team → New agent (same name + visibility form)
One-screen overview of your install first-tree status
Send a message to another agent first-tree chat send <agent> "..."
Stop the daemon and sign out first-tree logout
Switch this computer to another account first-tree login <code> with the new user's connect code, then confirm the switch
Update the CLI in place first-tree upgrade