ChatGPT Dots vs Cursor Projects (OpenAI Dots)

Snapshot

ChatGPT Dots (OpenAI Dots) and Cursor Projects are both always-on coordinators with a cloud computer, shared context that grows over time, and ways to keep work moving without a fresh chat every time. The practical fork is what the work is for and where you live: Dots is a personal always-on agent inside ChatGPT that spans inbox, research, docs, and can hand coding into Work/Codex; Cursor Projects is a repo-centric coordinator inside Cursor that plans large engineering bodies of work and delegates coding to many subagents.

For other always-on comparisons, see Dots vs Grok Bot · Dots vs Meta Muse · Dots vs Claude · Dots vs Custom GPTs / Projects · Dots vs Codex vs ChatGPT Work. Dots setup: getting started · features.

Product framing

• ChatGPT Dots — Always-on agent in ChatGPT; you give a responsibility, sources, and a definition of done; it keeps making progress between conversations and asks when a decision needs you. Official materials: GPT-6 Astra, per-Dot cloud computer + browser, plugins into thousands of apps, Slack/Teams messaging, optional connect of one personal computer, can start or manage Work and Codex tasks.
• Cursor Projects — Coordinator agent inside Cursor for larger bodies of work (a feature, a migration, a full app, or ongoing “gardening”). The coordinator does not write code itself; it plans, delegates to coding subagents (cloud by default, local when needed), maintains shared Project context across machines, and can act on subscriptions (schedule, Slack, PRs) without a fresh prompt.

Indie write-ups (not official product docs) often frame the same split as “personal cross-app agent” vs “cockpit for engineering agent fleets.” Treat those as framing only; product facts below follow OpenAI and Cursor primary pages.

Where you work day to day

• Dots: ChatGPT desktop app or desktop browser (chatgpt.com/dots); after setup, the same Dot in the mobile app when supported (mobile web is not supported for creation). Message or call in ChatGPT; also Slack and Microsoft Teams. Context carries across those channels while messages stay in their respective surfaces.
• Cursor Projects: Inside the Cursor IDE. Start a Project from the left-hand nav, chat with the coordinator, and review finished work the subagents bring back. Best when you already live in Cursor and the unit of success is branches, PRs, CI, and repo context.

If ChatGPT (and Slack/Teams around it) is already home for knowledge work, Dots inherits that surface. If your day is PRs and migrations in Cursor, Projects is the native coordinator.

One coordinator vs many coding agents

• Dots: One primary Dot is the usual model today. It can run several responsibilities and use background agents / Work / Codex tasks in parallel, but the product story is “your Dot” as the personal coordinator. OpenAI says teams of Dots are envisioned over time; specialist Dots with their own identity are in focused enterprise pilots—not the default personal ship.
• Cursor Projects: Explicitly a fleet model. You direct one coordinator; it spins up as many coding subagents as the work needs, mostly in the cloud, with shared Project files so later agents reuse what earlier ones learned (how to test a service, preferred patterns, artifacts).

Pick Dots when the “agent” is an extension of you across apps. Pick Projects when the “agent” is a project lead for a codebase.

Computer, context, and local access

• Dots: Per-Dot cloud computer and browser. Optional connect of one personal computer (online + ChatGPT app open). Cloud sessions stay separate from your personal browser; site sign-in uses private request or Take over. Plugin permissions are shared with ChatGPT / Work / Codex within what you granted.
• Cursor Projects: Project runs on its own cloud computer so closing the laptop does not stop it. Shared Project context syncs across cloud and local machines its agents use. When something needs testing on your machine, the coordinator can start a local agent there. Source control must be connected for cloud agents; environment setup (snapshot / Dockerfile / agent-led) matters as much as the prompt.

Both keep working with the laptop closed for cloud work. Neither replaces human review on consequential changes.

Continuity and automation

• Dots: Continuity between chats is the default. Assign a responsibility; the Dot tracks progress, can decide when to follow up, and supports scheduled check-ins (time zone + end date). Proactive research on connected apps is described as read-only; consequential follow-ups still need permission. Activity View covers delegated Work/Codex tasks; Pause / Scheduled stop controls apply.
• Cursor Projects: Built for work that outlives a single chat. Subscriptions let the coordinator watch Slack, run on a schedule, or follow PRs (for example fixing CI or acting on open/merge)—so gardening and migrations can keep moving without you prompting each batch.

Approvals and control (public docs)

• Dots: Built-in safeguards, Auto-review against instructions / Custom Rules / safety requirements, plugin and computer permissions, Pause and Activity controls. Sensitive steps (for example changing a password) stay with you. OpenAI states Dots can still make mistakes—review consequential work.
• Cursor Projects: You stay in the review loop for finished PRs and results the coordinator brings back. Early migration PRs are meant to be reviewed closely; as the pattern holds, review can ease—but less review is a discipline choice, not a guarantee the fleet is correct. Cloud/subagent usage is metered; set spend limits where Cursor exposes them.

Pricing and access (as published)

• Dots: Rolling out to eligible Pro tiers (plan/region gates apply; Learn lists Pro 100/200/500 with age and market notes), Business Premium, and Enterprise when a workspace admin enables Use dots (Beta). First Dot included with eligible plans; Dot conversations do not count toward ChatGPT message limits; Work/Codex metering applies when the Dot starts or manages those tasks. Details: pricing · availability.
• Cursor Projects: In beta and rolling out to Cursor users; no separate sticker price called out on the launch post. Projects run on Cursor Cloud Agents / API-priced model usage under a paid Cursor plan with source control connected. Parallel subagents multiply token use—treat spend limits as part of the workflow.

Re-check both vendors’ current plan pages before buying or rolling out; this page does not replace them.

When to pick which

• Pick ChatGPT Dots (OpenAI Dots) if you want an always-on personal agent for cross-app knowledge work—inbox, research, docs, launch materials, ongoing personal or team responsibilities—inside ChatGPT, with optional handoff into Work/Codex for coding pieces.
• Pick Cursor Projects if the job is repo-centric multi-PR engineering (features, migrations across many files, design-system gardening, CI/PR subscriptions) and you already direct work from Cursor.
• Use both when that matches your stack: Dot for the surrounding product/ops loop; a Project for the long-running code body. They are not the same product surface—do not expect one to replace the other.

Related: Dots vs Grok Bot · Dots vs Meta Muse · Dots vs Claude · Dots vs Custom GPTs / Projects · Dots vs Codex vs ChatGPT Work.

Sources

OpenAI Introducing dots, Meet dots (Learn), Getting started / Help, Dots feature page, computers/apps and controls Learn pages, Work/Codex usage notes; Cursor Introducing Projects. Indie framing only: eesel Cursor Projects overview, CellCog OpenAI Dots notes—label as non-official. Re-check before buying or rolling out.