How to Use ChatGPT Dots as an Always-On AI Assistant
A practical guide to setting up ChatGPT Dots for proactive email monitoring, Slack alerts, and recurring research tasks.
What is ChatGPT Dots and how do you start using it?
Dots is OpenAI’s always-on assistant feature inside ChatGPT, introduced at the company’s dev day. Instead of juggling separate chats and projects, you talk to a single persistent entity, your dot, which sits above everything else in your account and decides on its own whether to pull in a chat model, Codex, web browsing, or a connected app like Gmail or Slack to get a task done. You access it from the same panel where you’d normally start a new chat. You can type to it or call it, the way you’d make a phone call. There’s no model picker and no need to decide which tool to use. You describe the outcome you want, and the dot figures out the rest.
TL;DR
- Dots replaces the chat/project juggling act by acting as one standing conversation that can reach into every tool connected to your ChatGPT account.
- It can run recurring scans, like checking rental listings every 30 minutes and only notifying you when something new matches your criteria.
- It’s proactive by default, flagging urgent emails or Slack messages, warning about a delayed flight, or compiling a “personal scratchpad” summary of your day without being asked.
- Setup is fastest if you already have connectors enabled, since Dots immediately inherits access to Notion, Todoist, Slack, GitHub, Granola, Google Calendar, Google Drive, Gmail, and iMessage.
- Access is currently gated by price, requiring a Pro plan (the cheapest tier that includes it) or higher, which puts it well above free always-on competitors like Meta’s Muse.
- Scheduled tasks created through Dots show up in ChatGPT’s task scheduler, where you can inspect, edit, or change the run interval manually.
Seven tools to build an app. Or just Remy.
Editor, preview, AI agents, deploy — all in one tab. Nothing to install.
How do you set up a recurring monitoring task in Dots?
The core workflow is simple: describe the task, set the cadence, and let the dot handle execution. A demonstrated example from a creator testing the feature: ask the dot to scan rental listings in a specific city, within a price range, with minimum bedroom and parking requirements, and then tell it to re-run that search every 30 minutes and alert you only when a new listing matches. The dot uses OpenAI’s computer-use capability to actually browse listing sites, not just search a cached index, which matters for tasks where fresh information is the whole point.
Once it returns results, you can refine the criteria conversationally. In the demo, adding “I want a pool in the backyard” was enough for the dot to update the running scan without having to rebuild the task from scratch. That updated task then appears under ChatGPT’s scheduled tasks list, fully visible, where you can see the exact prompt it generated for itself and adjust how often it runs.
This same pattern extends to other recurring needs: tracking flight or hotel prices for an upcoming trip, watching a marketplace for a specific product, or monitoring LinkedIn for new job postings in a given role. The point isn’t a single clever use case, it’s that any task you’d normally repeat manually (check, compare, filter, notify) can be handed off wholesale.
How does Dots monitor email and Slack proactively?
Beyond tasks you explicitly assign, Dots is designed to surface things before you ask. In testing covered by a creator who got early access, the dot would flag incoming emails it judged important and offer to draft a reply. Connected to Slack, it did the same thing for urgent-looking messages, offering to craft a response rather than waiting to be asked.
It can also be instructed to triage an entire inbox: told to look through messages and separate what can be archived from what needs a response, it identified emails safe to ignore and drafted replies for the rest, without sending anything, respecting an explicit instruction to hold all sends for manual review.
The proactive behavior went further over a week of testing. The dot flagged an upcoming meeting and related email unprompted on a Monday, surfaced a relevant Slack thread on its own the following day, and caught a flight delay before the traveler saw the airline’s own notification. On a day with multiple commitments, it compiled a “personal scratchpad,” a single document pulling together emails, Slack messages, and calendar context into one agenda, so there was one place to check instead of several apps.
What connectors and tools does Dots actually use?
One coffee. One working app.
You bring the idea. Remy manages the project.
Dots doesn’t operate in isolation. It draws on whatever you’ve already connected to your ChatGPT account. In the dev day demo coverage, that included Notion, Todoist, Slack, GitHub, Granola, Google Calendar, Google Drive, Gmail, and iMessage. Anything already linked becomes immediately available to the dot without additional setup, which is a meaningful part of the onboarding story: if you’ve connected your tools to ChatGPT already, Dots doesn’t require you to reconfigure anything.
For technical tasks, Dots can route work to Codex, OpenAI’s coding agent, rather than making you decide to open a separate Codex session. Ask it to build something, and it picks whatever underlying model or tool it judges best for the job. This “no model selector” design is the main interface idea behind Dots: you stop thinking about which product inside ChatGPT you need and just state the goal.
Is ChatGPT Dots worth the price?
This is where Dots gets harder to recommend universally. Access currently requires a Pro plan, the entry point for using the feature at all, with Business and Enterprise tiers also eligible, and availability itself is limited to certain markets. OpenAI also reintroduced a higher-usage tier above that and added a new, more expensive tier on top for people who need heavier usage, including access to an accelerated “ultra fast” inference mode that trades a steep price multiplier for significantly higher token throughput.
That pricing puts Dots well above free alternatives. Meta’s Muse, a comparable always-on assistant, is available at no extra cost to anyone with a Meta account and already supports a wide range of connectors, though it lacks a few integrations Dots has, like Granola and direct iMessage access. Grok’s equivalent assistant rounds out a now-crowded field of “always-on agent” products competing on the same basic premise.
The honest tradeoff: Dots runs on OpenAI’s strongest models and benefits from deeper, more mature connector support and tighter integration with Codex for coding tasks, which makes it noticeably more capable in side-by-side testing. Whether that capability gap justifies the price difference over a free tool depends heavily on how much of your day already lives inside OpenAI’s ecosystem versus Meta’s or Google’s.
Frequently Asked Questions
What exactly is a “dot” in ChatGPT?
It’s a persistent, always-on assistant that sits above your regular chats and projects, with access to your connected apps and the ability to decide which underlying model or tool (chat, Codex, browsing) to use for a given task, without you having to choose manually.
Can Dots monitor things in the background without me asking each time?
Yes. Once you give it a recurring task, like scanning listings every 30 minutes, it continues running on that schedule and only notifies you when results change. It also proactively flags things like urgent emails, Slack messages, and schedule changes on its own.
Do I need to reconnect all my apps to use Dots?
No. Dots inherits whatever connectors are already active on your ChatGPT account, including tools like Gmail, Slack, Google Calendar, Notion, and GitHub, so if you’ve already linked those, the dot can use them immediately.
How is Dots different from using ChatGPT projects?
Projects group related chats so they share context, but you still have to manage multiple projects and chats separately. Dots sits above all of that as one single conversation that can pull from every chat, project, and connected tool in your account.
Which plan do I need to use Dots?
Remy doesn't write the code. It manages the agents who do.
Remy runs the project. The specialists do the work. You work with the PM, not the implementers.
It currently requires a Pro-tier ChatGPT plan at minimum, with higher-usage and Business/Enterprise tiers also supported, and it’s only available in select markets so far.
