Cursor Projects: how the coordinator agent plans and delegates larger work

What Cursor Projects is

• A coordinator and its workers: you talk to one coordinator agent, which creates worker agents, names them by job and manages them. The coordinator doesn't write code itself; it plans the work and delegates it to the workers, which run in parallel.
• It runs in the cloud: each Project has its own cloud computer on Cursor's Cloud Agents, so closing your laptop doesn't stop it.
• Shared files: a Project keeps preferences, notes and workflows that every agent in it reads. A shared notes.md tracks what's in progress, queued and done.
• A task list and board: each Project has one, and in Multiplayer Projects every member shares it.
• Recurring work: tell the coordinator to watch a Slack channel, run on a schedule or follow your pull requests, and it acts on those signals without a prompt.
• Built for work that outlives one chat: a feature with several pull requests, a migration, or a job handled while you're away.

Cursor Projects pricing: which plans include it

Projects run on Cloud Agents, so they're on every Cursor plan that has Cloud Agents. Prices from Cursor's pricing help page on November 13, 2026. Cursor's paid plans include a usage pool for Cursor models (Grok and Composer).

How to set up Cursor Projects

1. Check your plan and privacy setting

   You need a plan with Cloud Agents: Premium or higher, Teams, Enterprise (Cursor 3.21.9 or later) or a legacy paid plan. Projects aren't available with Privacy Mode (Legacy).

2. Open the Agents Window

   Projects live in the Agents Window in Cursor and run in the cloud.

3. Describe the larger job

   Tell the coordinator what to build, such as a feature, a migration or a full app, in your GitHub repositories. You pick the coordinator's model.

4. Add shared files

   Write down preferences, notes and workflows for the Project; every agent in it reads them, and notes.md keeps track of what's in progress, queued and done.

5. Check the work and make it recurring

   The coordinator brings finished work back for you to check. Ask it to watch a Slack channel, run on a schedule or follow your pull requests to keep going without a prompt.

Source: Cursor docs, Projects

How Cursor Projects keeps you in control

• Review: the coordinator brings finished work back for you to check.
• Visibility: the task list and board, and the shared notes.md, show what's in progress, queued and done.
• Separation: each Project runs on its own cloud computer. The coordinator starts a local agent when something must be tested on your machine.
• Teams: in Multiplayer Projects (team account, Cursor 3.26.13 or later), each member has their own coordinator, the coordinators message each other directly, and every member shares the board.
• Privacy: Projects aren't available with Privacy Mode (Legacy).

Can Cursor Projects run locally?

No, not end to end. Cursor's docs say a Project runs on its own computer in the cloud, so closing your laptop doesn't stop it, and the cloud lets it run more agents in parallel than a laptop could. When something needs testing on your machine, the coordinator starts a local agent to run it there.

Cursor's Self-Hosted Machines move tool execution to a machine you manage: My Machines for one person's laptop, devbox or VM, and Team Pools on the Enterprise plan. File edits, terminal commands and other tool calls then run on that machine, but the agent loop, inference and planning stay in Cursor's cloud, and the worker sends Cursor the content the agent needs. (Cursor docs, checked November 13, 2026.)

If you want a whole team of agents running on your own computer, these work that way:

• Claude Code agent teams: an experimental Claude Code feature, off by default, where a lead session spawns teammates that share a task list and message each other, all in your own Claude Code sessions.
• Claude Squad: an open-source (AGPL-3.0) terminal app that runs several coding agents at once, each in its own tmux session and git worktree.
• Vibe Kanban: an open-source (Apache-2.0) kanban board for coding agents that runs locally in your browser or as a desktop app; its community has maintained it since its company shut down in April 2026.
• Crewmark: open source (GPL-3.0). Each AI employee is a Claude Code or Codex conversation on your own computer, with a role and goals, and a heartbeat wakes it to take the next issue from a shared board. The work isn't limited to code: roles such as SEO specialist run the same way, and the web app can be self-hosted.

Cursor Projects vs Grok Bot and Claude Code

Facts from each vendor's own docs and pricing pages, checked November 13, 2026. More AI coworkers are compared in AI coworker tools; see also the Grok Bot guide and Grok Bot vs Claude Code.

Side by side: Cursor Projects vs Grok Bot. Guides: Grok Bot.

Projects for one big job. Crewmark for standing roles.

Cursor Projects coordinates agents on a larger body of work. Crewmark turns your own Claude Code or Codex sessions into AI employees with a role, goals in numbers, working hours and a heartbeat, each taking issues from a shared board, so engineering, SEO and marketing work in a repository keeps getting done.

1. Pick a role

   Start from the AI employee roles, such as a software engineer or an SEO specialist.

2. Put the work on a board

   Assign issues and make the regular ones repeat; the heartbeat wakes the employee to take the next one.

3. See what it can do

   Jobs with setup are in Claude Code use cases and Codex use cases.

Cursor Projects FAQ