Quickstart
Go from a fresh ORG-2 install to your first agent session, watched live and replayed, in about ten minutes.
This walkthrough takes you from a fresh install to a finished agent session you can replay event by event. It assumes ORG-2 is already installed — if not, start with Install ORG-2. You will need one of the following: an API key from a model provider, or an existing subscription to an agent CLI such as Claude Code, Codex, or Gemini CLI.
1. Open ORG-2 and decide how to sign in
The app opens on a sign-in screen. You have two paths:
- Login — sign in to an ORG-2 account, which unlocks the hosted and marketplace features.
- Continue without signing in — run entirely on your own keys. Nothing on this page requires an account.
Pick whichever fits. The choice sticks until you sign in or sign out, and you can change it later.
2. Complete the setup walkthrough
The first-run wizard lists six steps down the left edge. Use Continue and Back to move, click any step to jump to it, or press Skip Setup to leave early — nothing here is mandatory.
- Welcome — "Let's configure your workspace in a few quick steps."
- Theme — choose Light or Dark. Changeable later in Settings.
- Set up Dev Passport — pick a Code Name and an Avatar, then Generate Vibe Passport. This is the identity your work is attributed to inside ORG-2.
- Connect GitHub — press Connect GitHub Account to reach repos, commits, and pull requests from inside the app. You can also connect it later from Settings → Integrations.
- Open a Workspace — covered in the next step below.
- Complete — press Get Started.
Finishing (or skipping) drops you into the Workstation, the main working surface.
3. Connect a repo
At the Open a Workspace step, choose one:
- Open Folder — point at a local git repo or any folder you already have.
- Clone Repo — clone from GitHub. The clone form offers a My GitHub tab (your repos, once GitHub is connected) and a GitHub URL tab, plus a Clone to parent folder picker.
If you skipped the step, add a workspace later from the sidebar's Folders section using Add workspace. Multiple repos can be grouped into a Multi-Repo Workspace (2–5 repos) when a task spans services.
4. Add a provider key or sign into an agent CLI
Open Settings → Models & Keys. The page is titled Models & CLIs and keys are stored locally — they never leave your machine.
For a model API key: stay on the My Keys tab and press Add Account. The wizard asks for a Provider, then a Connection method — API Key or Subscription. Paste the key and let it validate; the account row moves to Ready to use when the check passes.
For an agent CLI subscription: switch to the CLI Clients tab and press Add Agent. Several CLIs support a browser sign-in instead of a pasted key, for example Sign in with Claude Code, Sign in with Codex, and Sign in with Gemini. ORG-2 captures the authorization callback and stores the OAuth token locally. Each CLI row also shows whether it is Installed, its key status, and its install script.
You only need one working account to continue. See API keys for the full picture.
5. Start your first session
In the left sidebar, click New Session.
- Choose a Session category — the picker groups agents under Built-in Agents (such as SDE Agent for coding and OS Agent for computer-use automation), CLI Agents ("Connect to CLI-based coding agents"), Custom Agents, Agent Teams, and External Agent.
- Choose a Model and, where relevant, an Account — these are the keys you added in step 4. The selectors are labeled Select model and Switch account.
- Type your task in the prompt box. The placeholder tells you the two shortcuts that matter: "Type a message, @ to add files, / to use skills." Use
@to pull specific files into context and/to reach skills, modes, and models. - Press the send arrow (Send).
Keep the first task small and verifiable — "add a test for X", "rename Y across the repo", "explain how Z works". If your prompt is very short, ORG-2 asks you to confirm before launching.
6. Watch it run
The session opens in the Workstation and streams as it works. Two things to look at:
- The trajectory. The bottom panel has a Trajectory tab (with a Todo tab beside it) that fills with events as they happen — tool calls, file reads and edits, terminal commands, thinking. Before the first event lands it reads "No events yet".
- The workstation tabs. As the agent opens files, runs commands, or drives a browser, tabs appear in the dock. The filter above them switches between All Tabs, Code, Browser, Database, and Projects.
You do not have to sit and watch. Sessions are long-running; leave and come back.
7. Review what happened
When a turn finishes, the chat collapses it into a summary line — "Agent worked for …" — with the number of tool calls and files changed. From there:
- Press Review on the changed-files strip to read the diff for that round.
- Use View Diff to open the full session diff in its own window.
- Open Ops Control from the sidebar for the board view of every session, with columns like Todo, In Progress, Your Turn, and Finished. The view switcher also offers List and Diary.
The session detail view surfaces the repo, branch, round count, models used, files changed, and Event Analytics — tool usage, token usage, errors, and duration.
8. Replay it
Replay is the reason the trajectory is recorded in the first place: it reconstructs what the agent saw and did, in order, instead of leaving you with a diff and a wall of text.
- Whole session: select the session in Ops Control and press Replay in the detail header. Playback starts from the first event.
- Single turn: in chat, use Replay turn → Play replay on a completed turn.
During playback you get Play / Pause, Previous event / Next event, and Playback speed. Filter events narrows the stream to Key interactions, File changes, Terminal events, Explore, or Other. The view either follows the agent (Follow Agent) or lets you wander independently (Free browse) while the timeline keeps running.
Next steps
Questions? Ask in the ORG-2 Discord. Discord.