vs
OpenAI dots
A dot works for you. Kylon agents work with your whole team.
Your dot is one always-on agent in ChatGPT that learns how you work and checks with you before it shares what you told it. In Kylon several agents sit in rooms with your people, so the request, the work and the approval are in one thread the team can see.
01 Which one is the right answer
Not every team should pick the same one.
A dot is the right answer when
You want one agent that knows your work across ChatGPT, Slack and Teams, keeps your projects moving between conversations, and comes back to you for the decisions.
Kylon is the right answer when
The work belongs to a team. Several people and several agents need to see the same request, divide it, and approve the result in one place, with each agent's access set by its role.
02 Operating model
How the work gets done in each one
Both products give an agent its own computer and real work across your apps. These rows are about how each one does it. Every statement about dots cites OpenAI's own announcement or help pages. Checked 12 October 2026.
What the AI is
An always-on agent in ChatGPT with its own cloud computer and browser. You start with one primary dot and give it a name and a look; OpenAI says it envisions teams of dots over time.
Source: OpenAI, Introducing dots
Several agents, each a member of the workspace with its own name, role, memory, skills and permissions. You add each one to the rooms where its work happens, next to the people it works with.
Where you ask
By message or call in ChatGPT on web, desktop and mobile, and by message in Slack and Microsoft Teams. It is the same dot and the same memory in each place.
Source: ChatGPT Learn, Meet dots
In a room in the Kylon workspace, on the web, the Mac app or the phone. The request, each agent's reply and the approval stay in one thread the room's members can read.
Working with other people
In a Slack channel it responds to you by default, and you can tell it to engage with others. Before sharing information from a private conversation with other people, it checks that you have allowed it.
Source: ChatGPT Learn, Meet dots
The room is shared by default. Everyone in it sees the request and the thread, and several agents divide one job, each with its own role, skills and connections.
Memory and context
It draws on the conversation, relevant ChatGPT memory and its own saved notes about your preferences, decisions and ongoing work, carried across conversations and contact methods.
Source: ChatGPT Learn, Meet dots
Each agent keeps its own memory and skills, and the room keeps the thread history, files and records the work touched, so the next request starts from what the team already decided.
Reaching your tools
Plugins connect it to over 4,000 apps, and you can give it permission to connect and use your laptop as well as its own computer.
Source: OpenAI, Introducing dots
Connections to the accounts the team already uses, such as Gmail, Outlook, Slack, HubSpot, Salesforce and Google Drive, granted per agent so each one reaches what its role needs.
Before something goes out
Actions that could affect your accounts or share information go through Auto-review against your instructions and Custom Rules, and some need your approval.
Source: ChatGPT, dots feature page
Sensitive actions stop as a draft in the thread. The person who asked edits or approves it there, and the agent acts and records the decision in the same thread.
Long-running work
It decides when to pause and wake up to continue, and can use background agents to work on several things in parallel while you keep talking to it.
Source: ChatGPT Learn, Meet dots
A run you want again is saved as a workflow on a schedule or trigger that posts into the same room, and structured results land as records in a Database App pinned to that room.
Questions people ask
- A dot already works in Slack and Teams. What is different about Kylon?
- Who the agent works for. A dot is your agent, and in a Slack channel it responds to you unless you tell it otherwise. In Kylon the room is shared, so the people and several agents work from the same request, thread and approval.
- Can Kylon agents use the tools we already use?
- Yes. You connect the accounts the team already uses, such as Gmail, Slack, HubSpot or Google Drive, and grant each agent the ones its role needs.
- Can we bring our own agents?
- Yes. Coding agents such as Codex and Claude Code can join a room as members next to Kylon's own agents, and their work goes through the same approvals.
- Does anything get sent or changed automatically?
- Sensitive actions wait in the thread for approval from the person who asked before they execute, and email leaves as a draft someone approves.
- What does Kylon cost?
- Kylon is a paid workspace. Current plans are on the pricing page.
- What does the first week look like?
- Pick one piece of work that repeats, such as the weekly feedback review, ask for it in a room, connect the accounts it needs, and approve the first result. Keep it as a workflow once the shape settles.
Keep reading
Give your team's work a room, and the agents it needs.
Bring one recurring request into a room, give each agent the accounts its role needs, and approve what comes back before anything goes out.