AQ
DocumentedA browser-based, multiplayer workspace that runs agent CLIs on a team runner.
- Why it fits
- Current sources document parallel sessions, workspace isolation, visual review. Teams that need shared terminals, previews, review comments, and persistent remote sessions.
- Check first
- Execution depends on a connected self-managed or AQ-managed runner, and the GUI implementation is not public.
- Harness coverage
- Antigravity, Claude Code + 4 moreSix named agent CLIs plus plain shells. Each user signs in with their own CLI account or subscription; AQ does not resell agent usage.
Evidence and implementation record
Capability claims
- documentedparallel sessions
Multiple workspaces and agents can run concurrently.
- documentedworkspace isolation
Each workspace receives its own Git worktree and branch.
- documentedvisual review
The browser workspace includes an editor, live previews, comments, and PR state.
- documentedremote execution or access
Persistent sessions run on a runner VM and survive browser or laptop disconnects.
- documentedshared team access
Teammates can enter the same live terminal, editor, and preview.
First-party sources
- How AQ works
Worktrees, persistent sessions, supported agents, shared access, previews, and runners.
- AQ product surface
Browser workspace, team access, agent CLIs, previews, runner choices, and bring-your-own credentials.
- Parallel Claude Code sessions
Per-task worktrees, persistent tmux sessions, automatic setup, preview ports, collaboration, and cleanup.
- Persistent remote sessions
Runner custody, persistent terminals, remote reconnection, per-user authentication, worktrees, and shared steering.