A Cursor background agents alternative for teams that keep code on their own machines
For heads of software companies whose team uses Cursor. Cursor's background agents, now called Cloud Agents, run in Cursor's cloud VMs or on machines you register; mcptask.online is a task queue for people and AI developers, where the runner works on a PC or Mac the team sets aside, with the project's real database and tests, and ends each task as a pull request after tests and local CI.
What are Cursor background agents?
Everything below comes from Cursor's own documentation and pricing page, linked under the comparison table.
- Cursor's documentation says Cloud Agents were formerly called Background Agents: agents that run in isolated cloud VMs with a full development environment instead of on your laptop.
- A Cursor Cloud Agent is started from the Cursor desktop app, cursor.com/agents, the iOS app, Slack, a GitHub or Bitbucket comment, Linear or the API; it works on a separate branch and produces a merge-ready pull request with artifacts.
- Cursor now offers Self-Hosted Machines for Cloud Agents: Cursor still runs the agent loop, inference and planning, while your worker machine runs file edits, terminal commands and local MCP servers. Team Pools of such workers require the Enterprise plan.
- Cursor Cloud Agents use a curated selection of models and draw on your Cursor plan's usage at the selected model's rate; they are included in Premium and higher, Teams and Enterprise plans.
How is mcptask.online different?
- mcptask.online is the task queue itself: people and AI developers work from the same backlog, so an AI developer picks up work the way a colleague does instead of waiting for someone to start it.
- The mcptask runner runs on a PC or Mac you set aside (Linux and Windows work too), with your project cloned, against your real database and tests; the whole agent works there, not only its tools.
- The mcptask runner takes the next task from the team's queue itself: priority, then bugs first, then oldest.
- Every mcptask runner task ends as a pull request after tests and local CI on the runner machine. In auto-squash modes the runner merges only after green checks; in manual modes a person reviews every pull request.
- The mcptask runner drives Claude Code, Codex CLI or OpenCode, on your own subscription or API key, with the model of your choice, and logs its time on every task like a person does.
- mcptask.online prices AI developers as users, at the same price as people: one person and one AI developer are two users.
How do Cursor Cloud Agents and mcptask.online compare?
Cursor's column states only what Cursor's own docs and pricing page say, with the links below the table. “–” means we found no statement either way.
| Cursor Cloud Agents | mcptask.online | |
|---|---|---|
| Where it runs | Isolated VMs in Cursor's cloud; Self-Hosted Machines run tool execution on workers you manage while Cursor runs the agent loop | The mcptask runner on a PC or Mac you dedicate (Linux and Windows too), with your project, real database and tests |
| Where tasks come from | Started by a person from Cursor, cursor.com/agents, mobile, Slack, GitHub, Bitbucket, Linear or the API | The runner takes the next task from the team's queue itself: priority, then bugs first, then oldest |
| Coding CLI and model | Cursor's agent with a curated selection of models | Claude Code, Codex CLI or OpenCode on your own subscription or API key; the model is your choice |
| Before merge | A merge-ready pull request with screenshots, videos and logs | Tests and local CI, then a pull request; auto-squash modes merge only after green checks, manual modes wait for a person's review |
| Timesheets | – | AI developers log their work like people do; it lands in the same timesheet and time reports |
| Git hosts | GitHub, GitLab, Azure DevOps, Bitbucket Cloud | GitHub, GitLab (including self-managed) and Bitbucket Cloud |
| Pricing | Included in Premium and higher, Teams and Enterprise; usage at the selected model's rate | Per user; an AI developer is a user at the same price as a person |
Cursor sources
- https://cursor.com/docs/cloud-agent
- https://cursor.com/docs/cloud-agent/self-hosted
- https://cursor.com/pricing
Sources checked 2026-11-11.
When is Cursor the better choice?
- Cursor is the better choice when your developers already work in the Cursor editor and want to hand a task to an agent from there, from Slack or from a pull request comment.
- Cursor is the better choice when you want many agents in parallel on managed VMs, with no machine of your own to keep running.
- Cursor is the better choice when you want artifacts such as screenshots and videos in the pull request, or remote desktop control of the agent's machine.
Can I use both?
- Yes. Your developers can keep using Cursor and its Cloud Agents for the work they start themselves, while the mcptask runner works through the team's queue on its own machine.
- The runner's pull requests land in the same repository, so your review rules and checks apply to both.
- Start with one project and one runner machine; 30 days free, no card needed.
Frequently asked questions
Are Cursor background agents and Cursor Cloud Agents the same thing?
Yes. Cursor's documentation says Cloud Agents were formerly called Background Agents. They run in isolated cloud VMs, or on Self-Hosted Machines where your worker runs the tools and Cursor runs the agent loop.
Where does the mcptask.online AI developer run?
The mcptask runner runs on a PC or Mac you set aside (Linux and Windows work too), with your project cloned and your real database and tests. Task management, timesheets and the queue are the hosted mcptask.online service.
Who decides what the mcptask runner works on?
The team's task queue. The runner takes the next task itself: priority first, then bugs, then the oldest task, the same queue people on the team work from.
What happens before the runner's code is merged?
Every task ends as a pull request after tests and local CI. In auto-squash modes the runner merges only after green checks; in manual modes a person reviews every pull request.
How are AI developers priced on mcptask.online?
As users, at the same price as people. Each AI developer is one user, so one person and one AI developer are two users. Every plan starts with 30 days free.