Skip to content
NinebrainsNinebrains

First run

No account, no sign-in

Ninebrains has no account and no sign-in screen. The first time you open it, you go straight to the main window. Add a project from the sidebar to begin, as in Getting started.

Telemetry is off, and no telemetry endpoint is built into the app.

Importing Emdash data

If the app finds data from an earlier Emdash install that has not been imported yet, it shows an import step first:

  • Import data imports it. You choose which sources to import.
  • Start fresh skips it.

After an import, a welcome screen appears. Click Start shipping (or press ⌘Enter / Ctrl+Enter) and the app reloads.

Agents

Settings → Agents lists the agent CLIs the app found on your machine. Lanes offer only Claude Code and Codex, and only if they are installed. Log in to each CLI yourself, in a terminal. See Accounts.

Hooks in your agent settings

Status lights, notifications and resume come from the agents’ own hooks. The first time an agent session starts, Ninebrains checks that its hooks are installed and adds them if they are missing.

AgentFileHooks added
Claude Codesettings.json in ~/.claude, or in CLAUDE_CONFIG_DIR if setSessionStart, UserPromptSubmit, Notification, Stop
Codexconfig.toml in ~/.codex, or in CODEX_HOME if setSessionStart, PermissionRequest, Stop

These are your user-level settings, so the hooks are present in every Claude Code or Codex session you run, including ones outside Ninebrains. They do nothing there: each hook command exits at once unless Ninebrains started the session. Your own hooks and other settings are kept.

Settings → Agents shows the state in the Hooks row: “Configured on first session” before they are added, and “Configured” with the folder once they are.

Claude lanes also carry the same hooks in their own per-launch settings file.

The folder trust prompt

Claude Code asks whether you trust a folder the first time it runs in it. Every task and lane has its own worktree, so without help you would see this prompt for each one.

Settings → General → Auto-trust worktree directories answers it for you. It is on by default. When it is on, Ninebrains marks each new worktree as trusted in ~/.claude.json before the agent starts. When it is off, the agent asks as usual.

The trust record is per Claude config directory. See Accounts.

Worktrees

Every task and every lane gets its own git worktree. By default they are created under ~/ninebrains/worktrees, in one folder per repository. Lanes use branches named lanes/<id>.

GitHub

Open Settings → Integrations and connect GitHub. The Connect GitHub dialog offers:

  • Import from GitHub CLI. Uses the accounts you have already signed in to with gh. If there is no session, it tells you to run gh auth login first.
  • Use device flow. Sign in on this device with a one-time code. This option appears only in builds that carry a GitHub OAuth app ID. Builds without one, including v0.1 releases so far, offer only the GitHub CLI import.
Terminal window
gh auth login

Then choose Import from GitHub CLI.