AQ vs Warp Oz: agent runs you review vs a place your team works
Published July 10, 2026 · by the AQ team
Choose Warp Oz if you want to dispatch large fleets of agent runs to the cloud, route them across models, and review the results from one dashboard. Choose AQ if you want your team working inside live, shared agent workspaces: the same terminal, editor, and app preview, running on infrastructure you control. The one-line difference: Oz is oriented around agent runs that you launch and review, AQ is oriented around workspaces that humans and agents occupy together.
Both products exist because single-player agent tools stopped matching how teams actually work. Warp's own launch framing made the point crisply: "Claude Code is built for individuals, Oz is built for teams." AQ agrees with the diagnosis and takes a different cure, so the comparison is mostly about which team model fits yours.
AQ vs Warp Oz at a glance
| AQ | Warp Oz (as of July 2026) | |
|---|---|---|
| Core unit | A workspace: an isolated git worktree with live terminals, an editor, and a preview that people enter | An agent run: a task dispatched to a cloud agent, tracked to completion |
| Platform | Browser-based: macOS, Windows, Linux, and mobile browsers | Orchestration platform from Warp; runs viewed through dashboards and shareable links |
| Where agents run | Your own VM, or an AQ-managed dedicated VM (included with the paid plan) | Warp-hosted cloud or self-hosted |
| Supported agents | Claude Code, Codex, Cursor Agent, Kimi, Grok, plus plain shells, as real CLIs in tmux with each user's own logins | Multi-harness: Claude Code, Codex, and Warp's own agent, with routing across frontier and open-weight models |
| Team collaboration | Multiplayer workspaces: teammates see the same live terminal, editor, and preview and can type mid-run; share with specific people or the whole team | A single pane of glass across the team's agents, shareable run links, one-click session joins to audit agents, audit trails |
| Live app previews | Per-workspace dev server with shareable and guest links; comments pinned on the preview go to the agent as prompts | Not a stated feature of the Oz platform |
| Issue tracker intake | Label a Linear issue "ai-task" and a workspace appears in the team sidebar; status syncs both ways, ownership follows the assignee | Runs are dispatched from the Oz platform |
| Pricing | $100 per user per month, promotional launch pricing (standard $200); 14-day trial | Not published; Warp directs inquiries to sales |
What is Warp Oz?
Warp Oz (warp.dev/oz) launched on February 10, 2026 as "the orchestration platform for cloud agents," and Warp now frames it as "the most flexible way to build your software factory." It runs agent fleets in Warp's cloud or self-hosted, so you control the execution environment, compute, and network access if you want to. It is deliberately multi-harness: Claude Code, Codex, and Warp's own agent, with routing across frontier and open-weight models to balance cost and quality. For teams it offers a single pane of glass over every agent running across the org, shareable run links, one-click session joins for auditing, and audit trails. Warp also open-sourced its terminal core on April 28, 2026, which says something about the company's engineering credibility. As of July 2026, Oz pricing is not published; Warp points buyers at sales.
What is AQ?
AQ is the multiplayer workspace where engineering teams run AI coding agents like Claude Code and Codex together: shared live terminals, a code editor, and app previews, in your own cloud.
Every AQ workspace is an isolated git worktree on a VM you control (bring your own, or provision an AQ-managed dedicated VM (included with the paid plan) self-serve). Agents run as real CLIs (Claude Code, Codex, Cursor Agent, Kimi, Grok, or a plain shell) inside tmux, streamed live to the browser. Teammates who open the workspace see the same terminal scrollback, the same editor, and the same running preview, and they can type into the session mid-run. Each user signs into their own Claude, OpenAI, and GitHub accounts; AQ does not proxy or mark up model usage. Agents commit, push, and open PRs themselves, and AQ tracks each PR from open to merge per workspace.
The core difference: runs vs workspaces
Oz's primitive is the run. You dispatch a task, an agent executes it somewhere, and the artifact of collaboration is a link to that run: teammates can view it, join a session to audit it, and see it in the fleet dashboard. That model shines when the goal is throughput: many tasks, many agents, results reviewed after the fact.
AQ's primitive is the place. A workspace persists across many agent runs, human edits, preview sessions, and PRs. Teammates do not review a record of what an agent did; they walk into the room where it is happening, with the same live terminal and a running app in front of them. With AQ, the work is a place your team can enter, not a diff that arrives. If your bottleneck is dispatching more runs, Oz's model fits. If your bottleneck is humans steering, unblocking, and reviewing agents together, AQ's does.
Where Warp Oz wins
- Fleet-scale orchestration. Oz is built to launch and monitor many runs at once across a whole org, with audit trails, as of July 2026.
- Model routing. Routing across frontier and open-weight models to optimize cost is built into the platform. AQ leaves model choice to whichever CLI and subscription each user runs.
- A fully hosted option. Warp's cloud means zero infrastructure to stand up. AQ always runs on a VM you (or AQ, paid) provision for your team.
- Warp's pedigree. Years of terminal engineering, an open-sourced terminal core (April 2026), and a large existing developer audience.
Where AQ wins
- Live multiplayer, not just visibility. Oz's team features center on viewing and auditing runs. AQ teammates work inside the same session: same terminal, same editor, same preview, typing mid-run.
- Previews with feedback loops. Every workspace gets a live dev-server preview, shareable publicly and with guests outside the team; comments pinned on the preview (with screenshots and attachments) go to the agent as prompts.
- Your infra by default. AQ has no AQ-hosted execution tier at all: agents run on your VM or a dedicated VM provisioned for you, and CLI credentials are per user, never pooled through the vendor.
- No inference middleman. Agents use each user's own Claude and OpenAI subscriptions. AQ does not proxy or mark up model usage.
- Issue tracker as intake. Labeling a Linear issue "ai-task" creates a shared workspace with two-way status sync. Work starts where the team already plans it.
- Published pricing. $100 per user per month (promotional; standard $200) with a 14-day trial, versus talk-to-sales as of July 2026.
Pricing compared
Warp does not publish Oz pricing as of July 2026; enterprise inquiries go through sales, so budget it like an enterprise platform purchase. AQ publishes one number: $100 per user per month at promotional launch pricing (standard $200), billed as a team plan with a 14-day trial. On top of the seat price you pay for your own compute (your VM or an AQ-managed dedicated VM) and your own model subscriptions, with no markup from AQ.
Bottom line
Pick Warp Oz if you want a software-factory model: dispatch fleets of runs to Warp's cloud or your own, route across models, and audit results from one dashboard. Pick AQ if you want a shared-workspace model: agents and teammates in the same live terminals, editors, and previews, on infrastructure you control, at a published per-seat price. Both are honest answers to "agents for teams"; they just disagree about whether the team gathers around a dashboard of runs or inside the work itself.
Frequently asked questions
Is Warp Oz self-hosted or cloud only?
Both. As of July 2026, Oz runs agent fleets in Warp's cloud or self-hosted, and Warp states you control the execution environment, compute, and network access in the self-hosted case. AQ has no vendor-hosted execution tier: agents always run on your own VM or an AQ-managed dedicated VM (included with the paid plan).
Does Warp Oz support Claude Code and Codex?
Yes. As of July 2026, Oz is multi-harness: Claude Code, Codex, and Warp's own agent, with routing across frontier and open-weight models. AQ also runs Claude Code and Codex, plus Cursor Agent, Kimi, Grok, and plain shells, as real CLIs signed in with each user's own accounts.
How much does Warp Oz cost?
Warp does not publish Oz pricing as of July 2026; the site directs inquiries to sales. AQ costs $100 per user per month at promotional launch pricing (standard price $200), with a 14-day trial.
Can teammates work inside the same agent session in Warp Oz?
Oz offers shareable run links and one-click session joins for auditing agents as of July 2026, so teammates can see what an agent did or is doing. AQ goes further: a workspace is a persistent shared place where teammates see the same live terminal, editor, and app preview and can type into the session mid-run.
What is the real difference between Warp Oz and AQ?
Orientation. Oz is run-oriented: you dispatch agent runs to the cloud and review results through links and dashboards. AQ is workspace-oriented: humans and agents share a live workspace (terminal, editor, preview) on your own infrastructure, and collaboration means entering it, not viewing a record of it.