ChatGPT Dots: What They Can See, What They Can't, and Alternatives for Work (2026)

TL;DR

• ChatGPT dots are OpenAI's always-on agents, launched at DevDay on 29 September 2026. Each dot runs on GPT-6 Astra with its own cloud computer and keeps working on tasks between conversations.
• Dots are the most business-ready of the new proactive agents: available on ChatGPT Business Premium seats, in beta for Enterprise once an admin enables them, and reachable in ChatGPT, Slack and Microsoft Teams.
• The limits for company use are scope and controls. A dot works from the plugins one person connects, its memory is that person's memory, and the Enterprise beta does not yet support data or inference residency.
• Sentra is the company memory a dot can check with, joining ERP, CRM and HRIS with what people said in email, chat and meetings, and it acts first with a named person approving.

What are ChatGPT dots?

ChatGPT dots are persistent AI agents inside ChatGPT that work on ongoing tasks in the background, on their own cloud computer and browser, and bring results back for review. OpenAI introduced dots at DevDay on 29 September 2026 and describes each dot as powered by GPT-6 Astra (openai.com, learn.chatgpt.com).

A dot is a feature of ChatGPT, not a separate app. You reach it in the ChatGPT mobile and desktop apps or a desktop browser, by phone call through ChatGPT, and in Slack and Microsoft Teams. Mobile web is not supported (learn.chatgpt.com). It can pause and wake itself to continue work without a fixed schedule, and it can run background agents in parallel.

ChatGPT dots at a glance

Sources: learn.chatgpt.com, openai.com, help.openai.com.

ChatGPT dots pricing

OpenAI lists no separate price for dots. They come with the plans that include them (learn.chatgpt.com).

• Pro. ChatGPT Pro is offered at $100, $200 or $500 a month (help.openai.com).
• Business Premium. A Premium seat on ChatGPT Business costs $100 per user per month billed annually or $125 billed monthly, with five times the usage of a Standard seat (openai.com).
• Enterprise. Dots are in beta and must be switched on by an admin (learn.chatgpt.com).

On usage, conversations with your dot do not count toward ChatGPT usage limits, but tasks it starts or manages in Work or Codex count toward those products' limits. OpenAI says plans include an allowance for deeper work, with extended limits for the first month after launch (learn.chatgpt.com).

What ChatGPT dots do well

• Reach where people already work. A dot you can mention in a Slack channel or message in Teams fits company habits better than an agent in a separate app (learn.chatgpt.com).
• A real business path. Business Premium access and an admin-enabled Enterprise beta mean IT can decide who gets a dot, rather than employees signing up on personal accounts.
• Layered safeguards. Built-in safeguards, existing ChatGPT app permissions and automatic approval checks apply from the start, and custom rules cannot override built-in safety requirements. Research runs read-only (learn.chatgpt.com).
• Careful credential handling. Website sign-in uses a private form, credentials go to the browser outside the conversation, and saved logins need your confirmation before reuse (learn.chatgpt.com).
• Default training terms for business. OpenAI says it does not train its models on dots data from Business, Enterprise and Edu workspaces by default (help.openai.com).

What ChatGPT dots can see, and what they can't

A dot sees the plugins you connect. OpenAI's setup guide says to connect plugins for the apps you want your dot to use, such as email, calendar and files, and each must be connected and permitted for the action (learn.chatgpt.com). Connecting a messaging channel does not grant access to an inbox, other apps or your computer. That is good security, and it also means the dot's view of the company is one employee's view.

Its memory belongs to one person. A dot starts with relevant ChatGPT memory and keeps its own notes on your preferences, decisions and ongoing work (learn.chatgpt.com). It does not begin with the company's history: which pricing exception was approved last quarter, or who owns an account after a reorg.

Adding it to Slack is not monitoring Slack. OpenAI notes that adding a dot to a Slack channel does not start monitoring that channel (learn.chatgpt.com). Conversations it is not asked about stay outside its view.

Enterprise controls are still in beta. Dots do not support data or inference residency during the beta (learn.chatgpt.com). OpenAI's privacy FAQ says dots keep context for as long as the dot exists and that deleting the dot clears it, while files and conversations it created are stored separately (help.openai.com). Regulated buyers should check retention, logging and residency against their own policies before rollout.

Plugins return records; joining them is your problem. A dot can read a CRM through one plugin and email through another. OpenAI does not describe dots reconciling a system of record with the conversations around it, such as matching a stock figure in the ERP with the store manager's message saying the shelf is empty.

Stopping is not undoing. OpenAI says stopping a task does not undo actions already completed, and some websites block cloud browsers (learn.chatgpt.com).

These limits are typical of a first release, and OpenAI states most of them plainly. They define what a company still needs around the agent. Our guide to ChatGPT company knowledge covers the same gap in ChatGPT itself.

ChatGPT dots for business: how to roll them out

1. Start with Business Premium seats for a pilot group. Dots roll out on Premium seats, so a mixed workspace can give them to the people who need them first.

2. On Enterprise, decide the beta policy. Dots are off by default. Set who can enable them and which plugins they can use before you switch them on. OpenAI's admin docs say dots use admin-managed remote MCP hooks on the cloud orchestrator (learn.chatgpt.com).

3. Review residency and retention. If your contracts require data residency, the beta is not yet a fit.

4. Write the custom rules. Use custom rules for what a dot may do on its own, and keep payments, external email and record changes on approval.

5. Give every dot the same company context. Without a shared source of truth, ten employees' dots will give ten answers to the same question.

For the wider comparison between ChatGPT and other assistants, see ChatGPT vs Copilot vs Claude vs Gemini.

ChatGPT dots vs Claude Duet

Anthropic opened Claude Duet as a research preview on 21 October 2026, three weeks after dots. Both run on their own cloud computer and work in the background.

Sources: learn.chatgpt.com, support.claude.com.

For a company rollout today, dots has the clearer path. Our Claude Duet page covers Anthropic's agent in detail.

Works with Sentra

A dot knows its user. Sentra knows the company, and every dot can ask it.

Sentra reads what is happening across the company, in email, chat, meetings and documents and in ERP, CRM, HRIS and BI, and works out what is true now, why, and who owns it. It knows when a fact became true and when it stopped being true, so a dot asking for a customer's current terms gets the terms in force today. Answers respect each source's existing permissions, checked at query time, and a failed check counts as no access, so a dot acting for a sales rep sees what that rep could see.

Sentra exposes that memory over REST and MCP, so ChatGPT, Claude, Copilot, Gemini and internal agents use the same company memory. ChatGPT lets you add a custom MCP server from Plugins (help.openai.com), and dots use the plugins you install and enable. Connect Sentra wherever your dot supports MCP or custom connectors, under your workspace's plugin policy.

Sentra also acts first. It finds what is going wrong, works out who should act, and tells them, and a named person approves anything before it reaches anyone else. The Sentra Agent catches the promises people make in calls and threads and texts them before they are due.

Sentra is SOC 2 Type II and ISO 27001 certified, connects 200+ integrations, and runs in the cloud, in an isolated VPC, or air-gapped on premises. It replaces none of your tools.

Book a demo at /demo to see Sentra working under ChatGPT and your other agents.

ChatGPT dots alternatives

Sources: support.claude.com, town.com, sauna.ai, about.fb.com.

More options are compared in AI coworker platforms and personal AI agents at work.

How to choose

1. Match the agent to your main assistant. If the company runs on ChatGPT Business or Enterprise, dots is the natural pilot. Claude shops should watch for Claude Duet on Team and Enterprise.

2. Check your compliance needs first. Residency, retention and logging decide whether a beta agent can touch customer data.

3. Map where the truth lives. List the ERP, CRM, HRIS and BI systems behind your key numbers and ask how the agent reconciles them with what people said.

4. Run one task across agents. Give each the same real job, such as preparing a renewal brief, and compare what each one knew without being told.

5. Put shared memory under all of them. Expect several agents in use. A shared company memory keeps their answers consistent and permission-aware.

FAQ