AQ vs Cursor Cloud Agents: PRs that arrive vs a workspace you enter
Published July 10, 2026 · by the AQ team
Choose Cursor Cloud Agents if your team already lives in the Cursor editor and you want to fire tasks at agents from Cursor, Slack, Linear, or GitHub and get merge-ready PRs back. Choose AQ if you want agents running on infrastructure you control, inside live workspaces your whole team can enter while the agent is still working. The one-line difference: Cursor's cloud agents deliver results as diffs and PRs that arrive; AQ makes the work itself a shared place, with a live terminal, editor, and app preview your team occupies together.
These two products overlap on the promise (delegate coding work to agents that run somewhere else) and diverge on almost every mechanical detail: whose cloud, whose agent harness, whose model bill, and what teammates can see while a run is in flight.
AQ vs Cursor Cloud Agents at a glance
| AQ | Cursor Cloud Agents (as of July 2026) | |
|---|---|---|
| Platform | Browser-based: macOS, Windows, Linux, and mobile browsers; works with any editor | Cursor desktop, Cursor Web (cursor.com/agents), and the Cursor iOS app |
| Where agents run | Your own VM, or an AQ-managed dedicated VM (included with the paid plan); no vendor-hosted execution tier | A dedicated VM per agent in Cursor's cloud by default; self-hosted compute options exist, but the agent loop still runs in Cursor's cloud |
| Agent harness | Real vendor CLIs: Claude Code, Codex, Cursor Agent, Kimi, Grok, plus plain shells, in tmux with each user's own logins | Cursor's own agent, with a choice of underlying models |
| Team collaboration | Multiplayer workspaces: teammates open the same workspace and see the same live terminal, editor, and preview; shareable with specific people or the whole team | Per-person runs viewed from the dashboard; remote desktop control of the agent VM for testing |
| Previews | Live dev server per workspace, publicly shareable with guest links; comments pinned on the preview go to the agent as prompts | Artifacts attached to results: screenshots, videos, and logs |
| Issue tracker intake | Label a Linear issue "ai-task" and a shared workspace appears; status syncs both ways, ownership follows the assignee | Trigger agents from Slack, Linear, and GitHub or Bitbucket via mentions and comments |
| Results | A live session you can join mid-run, plus agent-opened PRs tracked per workspace from open to merge | Merge-ready PRs with attached artifacts |
| Pricing | $100 per user per month, promotional launch pricing (standard $200); 14-day trial; no markup on model usage | Cursor subscription plus cloud agent usage charged at API pricing for the selected model, with a spend limit |
What are Cursor Cloud Agents?
Cloud Agents are Cursor's remote agents, formerly called Background Agents. Each agent runs in an isolated VM with a full development environment in Cursor's cloud, so it can build, test, and interact with the software it is changing without touching your laptop. You can launch them from the Cursor desktop editor, from Cursor Web at cursor.com/agents, from the iOS app, by mentioning the agent from Slack, Linear, GitHub, or Bitbucket, or through an API. Results come back as merge-ready PRs with artifacts such as screenshots, videos, and logs, and Cursor offers remote desktop control of the agent's VM for hands-on testing. As of July 2026, teams can point agents at self-hosted compute pools, but Cursor's docs note the agent loop still runs in Cursor's cloud. Usage is billed at API pricing for the selected model, with a spend limit, alongside the Cursor subscription.
It is a polished system with an enormous installed base behind it, and if your engineers are in Cursor all day, the activation energy is close to zero.
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. Inside it, agents run as the actual vendor CLIs (Claude Code, Codex, Cursor Agent, Kimi, Grok, or a plain shell) in tmux, streamed live to the browser. Teammates open the same workspace and watch the same terminal, edit in the same editor, and click around the same running preview; workspaces can be shared with specific people or the whole team, and guest links let people outside the team view a preview. Agents commit, push, and open PRs with each user's own GitHub auth, and AQ tracks every PR per workspace. Model access comes from each user's own Claude and OpenAI subscriptions; AQ does not proxy or mark up usage.
Where Cursor Cloud Agents win
- Editor integration. Launching an agent is one gesture inside the tool your engineers already use, and the results land back in the same place. AQ works with any editor but is not embedded in one.
- Installed base and polish. Cursor is one of the most widely used AI editors, so cloud agents ride on accounts, billing, and habits that already exist.
- Trigger surface. Editor, web, iOS, Slack, Linear, GitHub, Bitbucket, and an API is a very broad set of entry points as of July 2026.
- Zero infrastructure by default. Cursor's cloud means nothing to provision. AQ requires a VM (yours or AQ-managed) before the first agent runs.
Where AQ wins
- Your infrastructure, actually. AQ agents run entirely on your VM or a dedicated VM provisioned for your team. With Cursor, even self-hosted compute keeps the agent loop in Cursor's cloud as of July 2026.
- Live and shared, not fire-and-forget. A Cursor cloud agent reports back when it has something to show. An AQ workspace can be entered mid-run by any teammate: watch the terminal, correct the agent's course, or take over the editor.
- One surface for the whole team. Cursor agents are launched per person from their editor. AQ workspaces live in a shared team sidebar, with visibility rules (private, shared with specific people, or team-wide) and ownership that follows the Linear assignee.
- Real previews, not screenshots. AQ runs the actual dev server per workspace and gives it a shareable URL; reviewers pin comments on the live app (with screenshots and attachments) and those comments go to the agent as prompts. Cursor attaches media artifacts to results.
- Any agent, your subscriptions. AQ runs the vendor CLIs themselves, including Cursor's own CLI, under each user's existing plan, with no per-token markup. Cursor cloud usage is metered at API pricing on top of the subscription.
The model difference: a diff that arrives vs a place you enter
Cursor's cloud agents are built around delegation: describe a task, receive a PR. That is exactly right for well-specified, reviewable chunks of work. It gets strained when the task is exploratory, when the agent goes sideways ten minutes in, or when three people need to see the same in-progress state. AQ's answer is that the work is a place your team can enter, not a diff that arrives: the run, the code, and the running app stay live in one shared workspace from the first prompt to the merged PR.
Pricing compared
As of July 2026, Cursor Cloud Agents are charged at API pricing for the selected model, governed by a spend limit you set, in addition to the Cursor subscription itself. AQ is $100 per user per month at promotional launch pricing (standard price $200), with a 14-day trial; compute is your own VM or an AQ-managed dedicated one, and model usage rides on each user's existing Claude or OpenAI plan with no markup from AQ. Which is cheaper depends on your usage: heavy agent workloads at API rates can dwarf a seat fee, while light usage may be cheaper metered.
Bottom line
If your team is all-in on the Cursor editor and wants delegated tasks returned as PRs with minimal setup, Cursor Cloud Agents are the path of least resistance. If you want agents on your own infrastructure, running any vendor's CLI under your own subscriptions, in live workspaces your team can watch, join, and steer together, choose AQ. The two are not mutually exclusive: plenty of teams keep Cursor as the editor and use AQ as the shared place where the team's agent work actually lives.
Frequently asked questions
Do Cursor Cloud Agents run on my own infrastructure?
Partially. By default each cloud agent runs on a dedicated VM in Cursor's cloud. As of July 2026 Cursor offers self-hosted compute options, but its docs note the agent loop still runs in Cursor's cloud. AQ runs everything (agent, terminal, editor, preview) on your own VM or an AQ-managed dedicated VM (included with the paid plan).
Are Cursor Cloud Agents the same as Background Agents?
Yes. Cursor renamed Background Agents to Cloud Agents; it is the same remote-agent product, with each agent running in an isolated VM in Cursor's cloud as of July 2026.
Can I run Claude Code with Cursor Cloud Agents?
No. Cursor's cloud agents run Cursor's own agent harness with a choice of underlying models, not third-party CLIs, as of July 2026. AQ runs the vendor CLIs themselves: Claude Code, Codex, Cursor Agent, Kimi, Grok, and plain shells, each signed in with the user's own account.
How are Cursor Cloud Agents billed?
As of July 2026, cloud agent usage is charged at API pricing for the selected model, with a spend limit you set when you start using them, on top of the Cursor subscription. AQ charges $100 per user per month (promotional; standard $200) and does not meter or mark up model usage: agents use each user's own Claude or OpenAI sign-in.
Can teammates watch a Cursor cloud agent while it works?
You can open runs from the dashboard and Cursor offers remote desktop control of the agent's VM for testing, as of July 2026. These are views of a run. An AQ workspace is one shared place: multiple teammates see the same live terminal, editor, and preview at once and can type into the session mid-run.