OpenAI Dots Multiple Agents: Running a Team of Dots - Features

OpenAI Dots Multiple Agents: Running a Team of Dots

Learn how OpenAI Dots multiple agents work: assigning goals, splitting tasks, managing permissions, and coordinating several dots at once.

2026-09-30
OpenAI Dots Wiki Team
Quick Guide
  • OpenAI Dots multiple agents let you run several always-on GPT-6 Astra agents side by side, each with its own goal
  • One dot per job: split research, monitoring, and communications into separate dots instead of overloading one
  • Each dot has its own dedicated cloud computer, browser, and connected app permissions
  • Autonomy settings control how independently each agent works between your check-ins
  • Workspace admins can restrict which members and capabilities are allowed in managed environments

Why Run Multiple OpenAI Dots Agents?

OpenAI Dots, announced at DevDay on September 29, 2026, are persistent AI assistants powered by GPT-6 Astra. Unlike a one-off ChatGPT conversation, each dot keeps working toward an assigned goal across sessions. Running multiple agents means you can parallelize: one dot handles ongoing research, another monitors your calendar, a third processes email follow-ups, all reporting back through ChatGPT, Slack, email, or SMS.

The strongest argument for multiple agents is separation of context. A dot built for competitive research should not also juggle your inbox. Focused goals give each dot a stable direction, cleaner permission scopes, and easier activity review.

Design Principle

Match each dot to a single responsibility, such as a project, workflow, or research topic. One focused dot outperforms one overloaded dot trying to do everything.

Use PatternSingle DotMultiple Dots
Ongoing researchOne broad topicSeparate dot per topic
CommunicationsMixed email and Slack workDedicated dot per channel scope
MonitoringSingle recurring checkParallel monitors per data source
Review effortOne large activity logSmaller, focused logs per dot

Structuring Your Agent Team

Before creating dots, map your recurring work into categories that benefit from an always-on agent rather than a single chat.

Research Dot

  • Collects and updates findings
  • Works across sessions
  • Uses cloud browser for web tasks

Operations Dot

  • Calendar review and reminders
  • Recurring monitoring checks
  • Proactive scheduled updates

Communications Dot

  • Email and Slack workflows
  • Follow-ups at the right time
  • Connected-app permissions only
Permission Hygiene

Only connect the apps each dot actually needs. A communications dot does not require file access, and a research dot rarely needs your Slack workspace. Narrow scopes reduce risk and simplify audits.

Step-by-Step: Setting Up Multiple Dots

Follow this workflow to launch a small team of dots from ChatGPT on desktop web or the supported desktop app.

1

Create and Name Each Dot

Open the Dots experience in ChatGPT and start a new dot per responsibility. Use short, descriptive names like "Market Research" or "Inbox Triage" so activity logs are easy to scan.

2

Define One Goal Per Dot

Write a clear ongoing outcome for each dot. A focused goal, rather than a list of unrelated prompts, keeps the agent's direction stable across work sessions.

3

Set Autonomy and Boundaries

Configure instructions describing what the dot should prioritize, what boundaries it must follow, and when it should ask for approval instead of acting.

4

Connect Apps Per Role

Attach only the relevant connections, such as Slack for team workflows, email for communications, or files for context. Connections follow user and workspace permissions.

5

Assign First Tasks and Schedule Work

Give each dot a concrete task tied to its goal, plus recurring or scheduled checks. Reminders and recurring summaries run between conversations.

6

Review Each Activity Dashboard

Open each dot's activity view to inspect completed actions, follow current tasks, and adjust configuration as workflows mature.

Coordination Tip

Use scheduled proactive updates so each dot reports at a fixed cadence. Stagger the schedules so you receive one digest at a time instead of simultaneous pings from every agent.

Permissions and Workspace Control with Multiple Agents

More agents means more connected surfaces. Dots permissions are configurable at both the user and workspace level, which matters when several dots run concurrently.

Control LayerScopeWhat It Governs
Connected app permissionsUser and workspaceWhich apps each dot may access
Cloud computer accessPer dotBrowser and multi-step computer tasks
Local computer accessOptionalAccess to your own device, separate from cloud
Autonomy settingsUser configurationHow independently each dot acts
Workspace availabilityAdministratorWhether members can use Dots at all
Enterprise controlsOrg adminsGovernance and restrictions across teams
Enterprise Note

Business and organizational workspaces can manage Dots availability through administrator settings, including which connected tools members' dots may use. Roll out multiple-agent workflows gradually within managed environments.

Cloud vs local execution stays relevant with multiple agents: each dot's dedicated cloud computer runs independently of your local machine, and you can open a dot's computer view to inspect or intervene in its session when manual review is useful.

Multi-Agent Launch Checklist

Before Going Live with Multiple Dots:

  • Define one clear goal and name per dot
  • Connect only the apps each dot needs
  • Set autonomy boundaries and approval triggers
  • Schedule staggered proactive updates
  • Bookmark each dot's activity dashboard for review

FAQ

Q: Can I run multiple OpenAI Dots at the same time?

Yes. Dots are designed as persistent agents you assign per responsibility. Running OpenAI Dots multiple agents in parallel works best when each dot has a single focused goal, its own connected apps, and its own activity dashboard for review.

Q: Does each dot get its own computer?

Each dot can use a dedicated cloud computer with its own browser environment, allowing multi-step work to continue independently of your local device. Optional local computer access is a separate, user-controlled capability.

Q: How do I keep multiple agents from duplicating work?

Assign non-overlapping goals, connect different app scopes per dot, and stagger scheduled updates. Review each activity dashboard regularly so you can catch redundant tasks early and adjust instructions.

Q: Can workspace administrators limit multiple-agent usage?

Yes. Workspace and Enterprise administrators control whether Dots are available to members and which connected capabilities are permitted, so multi-agent workflows can be governed at the organization level.