What are OpenAI dots?

OpenAI dots are always-on agents that OpenAI launched on September 29, 2026. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, connects to apps through plugins, and keeps working toward your goals between conversations.

How are OpenAI dots different from coding agents?

Dots are personal agents that act for one person across research, documents, planning, and some software work. Coding agents such as Droid are built around software work. They edit code in a repository, run commands and tests, and return changes for review.

Which companies ship personal agents?

As of October 2026, OpenAI ships dots, Meta ships Muse, Anthropic offers a research preview of Claude Duet, Grok Bot offers a team of always-on agents, and Google has announced Gemini Spark for the Gemini app. Plans, availability, and controls differ by vendor.

What changes when personal agents are used at work?

The organization has to decide whose identity each agent uses, which apps and data it can reach, which limits users cannot change, where the audit record lives, and which actions need a person to approve them.

Where does Factory fit next to personal agents?

Factory lets team members delegate engineering work to Droid from Slack, Linear, and Jira, and runs recurring work with Custom Automations. Sessions use a personal identity or an organization service account, can run on Factory-managed Droid Computers or on machines connected through Remote Control, and stay inside organization permission rules.

All articles

Enterprise AI

AI Coding Agents

OpenAI dots vs coding agents for engineering teams

September 29, 2026 - 4 minute read

OpenAI dots are always-on agents that work toward your goals across conversations. OpenAI introduced dots on September 29, 2026. Each dot runs on GPT-6 Astra, has its own cloud computer, and can connect to more than 4,000 apps through OpenAI's plugin ecosystem. You can reach your dot in ChatGPT, Slack, or Teams.

Dots belong to a wider category of personal agents, which act for one person with their own compute, connected apps, and background work. Engineering teams already use a different kind of agent. Coding agents such as Droid work inside repositories and hand back changes for review. Knowing where the two overlap helps a team decide which work goes to which agent.

How OpenAI dots work

A dot works on its own cloud computer, with its own browser and the apps you connect. You can open that computer at any time to see what the dot is doing, and you can let it connect to other devices, including your laptop.

Between conversations, a dot looks for ways to help through what OpenAI calls proactive research. That mode uses read-only tools over your connected apps, so it cannot send messages, change app content, or control your browser or computer. Acting on what it finds is a separate step governed by Custom Rules, which allow an action, require approval, or block it. Auto-review checks actions that could affect your accounts against your instructions, your rules, and OpenAI's safety requirements. Some tasks, such as changing a password, always stay with you.

One of OpenAI's examples is a coding workflow, and it is where dots meet the work coding agents do. A dot watches customer feedback, scopes small fixes, builds and tests them, and returns pull requests with videos of the changes.

OpenAI is also previewing specialist dots for organizations. Each one gets its own identity, credentials, and access to the systems it needs, and OpenAI says it is working with Microsoft to bring specialist dots under the governance controls in Agent 365.

Personal agents vs coding agents

A personal agent works for one person over time. It has a computer of its own, reaches the apps its owner connects, and keeps going after the conversation ends. As of October 2026, several vendors ship this pattern alongside OpenAI.

Meta Muse. Meta introduced Muse on September 8, 2026 as a personal AI agent that people message in the Muse app or in WhatsApp. It runs on a dedicated cloud machine, and a separate Sentinel agent must approve anything Muse does that reaches the internet.

Claude Duet. Anthropic's research preview of Claude Duet gives professionals and small business owners a proactive Claude with its own virtual machine. It sends a morning brief and waits for approval before it sends a message, makes a payment, signs a document, or deletes anything.

Grok Bot. Grok Bot launched on August 11, 2026 as a team of always-on agents that share one persistent cloud computer and can run saved workflows on a schedule.

Gemini Spark. Google announced Gemini Spark at I/O on May 19, 2026 as a 24/7 personal AI agent in the Gemini app.

Coding agents start from a different place. Droid, Claude Code, Codex, Cursor, and Devin are built around software work. They read a repository, edit code, run commands and tests, and hand back a change that a reviewer can inspect. The person directing them is usually an engineer, and the result lands in version control.

Because coding agents do the same kind of task, they can be measured head to head. In Factory's October 2026 Terminal-Bench 4 comparison, each harness ran every task three times on the same model as Droid. With Claude Opus 5.5, Droid passed 65.2% of tasks, against 62.1% for Cursor and OpenCode, 59.6% for Devin, and 57.6% for Claude Code. With GPT-6.1 Sol, Droid passed 57.6% against 54.0% for Codex. The Terminal-Bench comparison guide covers the method and its limits.

The practical split is scope, with dots spanning a person's workday and coding agents going deep on a codebase.

What changes when agents come to work

At work, the same agent can touch company repositories, customer data, and shared channels. Before personal agents spread beyond a few early users, the organization needs its own answers on five points.

Identity. An agent working from one person's private messages can reasonably use that person's credentials. Work that belongs to a team, such as answering requests in a shared channel or running a nightly job, should run under an identity the organization owns, so access keeps working when people change roles.

Permissions. A company needs limits that individuals cannot loosen, covering which tools an agent may use, which commands are blocked, and which plugins are allowed.

Data access. List the systems each agent can read and where it can send data, and check each vendor's training and retention terms for the plan your people use. OpenAI's dots announcement says it does not use content from ChatGPT Business, Enterprise, or Edu workspaces to improve its models by default.

Audit. Security teams need records across agents that show who changed an agent's access, which identity ran a session, and which settings applied.

Approvals. Approval rules should follow the company's existing process. Preparing a patch can be automatic while merging it still needs a reviewer.

Where Factory fits

Factory brings the delegated, always-available style of a personal agent to engineering work, with Droid as the coding agent and the organization setting the rules. Team members delegate work to Droid from Slack, Linear, or Jira. A Slack direct message runs as the person who sent it, with their own Git credentials and connectors, while a request in a shared channel runs as an organization service account.

For work that keeps coming back, Custom Automations start sessions from schedules, Slack messages, GitHub events, or webhooks, with webhook triggers in private preview. Sessions can run on a Droid Computer, a persistent machine that Factory hosts and that keeps files, packages, and services between sessions. Teams can also connect a machine they already manage with Remote Control. Organization permission rules and Enterprise Controls bound what any session can do, whichever model it uses.

A team can use both kinds of agent. A dot can keep one person's research, documents, and planning moving, while Droid takes the issues that should end in a reviewed pull request. The guide to running personal agents securely at work covers identity, compute, and permissions in detail, and OpenAI dots vs Factory compares the two topic by topic.

Frequently asked questions

What are OpenAI dots?: OpenAI dots are always-on agents that OpenAI launched on September 29, 2026. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, connects to apps through plugins, and keeps working toward your goals between conversations.
How are OpenAI dots different from coding agents?: Dots are personal agents that act for one person across research, documents, planning, and some software work. Coding agents such as Droid are built around software work. They edit code in a repository, run commands and tests, and return changes for review.
Which companies ship personal agents?: As of October 2026, OpenAI ships dots, Meta ships Muse, Anthropic offers a research preview of Claude Duet, Grok Bot offers a team of always-on agents, and Google has announced Gemini Spark for the Gemini app. Plans, availability, and controls differ by vendor.
What changes when personal agents are used at work?: The organization has to decide whose identity each agent uses, which apps and data it can reach, which limits users cannot change, where the audit record lives, and which actions need a person to approve them.
Where does Factory fit next to personal agents?: Factory lets team members delegate engineering work to Droid from Slack, Linear, and Jira, and runs recurring work with Custom Automations. Sessions use a personal identity or an organization service account, can run on Factory-managed Droid Computers or on machines connected through Remote Control, and stay inside organization permission rules.

Talk with Factory

Discuss your team’s software development, privacy, or deployment requirements.