ChatGPT dots memory starts with the memories in your ChatGPT account: your dot receives them and builds its own context from your conversations and connected apps, and OpenAI says you "currently cannot view, delete or directly modify individual dot memories" (OpenAI help). The only way to wipe what a dot has learned is to reset it, and disconnecting an app doesn't erase what it already took from there. We were impressed with our dot on day one and then stopped using it, and the memory model is part of why; that's below the mechanics. Docs checked 2026-10-21.
What are ChatGPT dots, and who can get one?
Your dot is "an always-on agent in ChatGPT that can help you take a project from an idea to work you can use" (OpenAI help). Dots were introduced on September 29, 2026, powered by GPT-6 Astra (OpenAI). Each dot works on its own cloud computer, connects to "over 4,000 apps" through plugins, can use your laptop with permission, and you message it in ChatGPT, Slack and Teams.
Who gets one: Pro, rolling out outside the EEA, Switzerland and the UK; Business Premium, in all supported regions; Enterprise, Edu and Healthcare, a beta that's off until an admin turns it on. The first dot is included on Pro and Business Premium (Pro starts at $100 a month), and for now you get one. If you can't find dots, OpenAI's checklist is: check the account and open chatgpt.com/dots, check eligibility (over 18, outside the EEA/CH/UK on Pro), ask your admin on Enterprise, and wait, since rollout "may take several days to reach your account."
What does your dot remember, and for how long?
For as long as you keep it: it "can retain context from your conversations and plugins for as long as you keep your dot" (dots FAQ). It starts with relevant ChatGPT memory and keeps its own notes on preferences, decisions and ongoing work, which "are separate from ChatGPT's saved memory and are not a complete transcript of everything you've said" (Tasks and memory). It can search your past ChatGPT conversations, it creates its own memories including from connected apps, and ChatGPT, Slack and Teams all reach the same dot and its notes, though the visible conversations stay separate. Each task has its own conversation and doesn't automatically receive every conversation you've had with your dot, which is the usual reason a dot seems to forget an instruction: give the correction in the affected task.
Can you see, edit or delete your dot's memory?
Not item by item. The FAQ is direct: you "currently cannot view, delete or directly modify individual dot memories." Three levers instead:
- Correct it in the conversation. Activity View in the desktop app lets you "add context, correct a misunderstanding, change direction, or ask a dot to stop" (OpenAI safety post).
- Disconnect an app. This stops new access, but "information a dot has already learned remains in its own context."
- Reset the dot. Profile, •••, Reset: that deletes its conversations, saved memories and scheduled tasks (OpenAI help). Your other ChatGPT memories stay.
Some people post that their dot recalls things they deleted or never saved in ChatGPT. The docs explain why: turning off ChatGPT Memory "does not delete information that your dot has already received," and the dot keeps its own notes. On training, OpenAI doesn't train on Business, Enterprise and Edu data by default; on personal plans the Improve the model for everyone setting decides.
What we found on the first call, and why we moved on
The first impression was strong. The thing that got us was being able to call our dot and just talk to it: on the first call it went and did a pile of research, looked things up, and then saved the findings and the follow-ups to our Tokenrip workspace, where the rest of our agents could pick them up. That was the first time a personal agent felt like a colleague instead of a chatbot. And because a dot has its own computer, it connected the way our Grok Bots do: we gave it the setup line, it installed the Tokenrip CLI on its machine and signed in, no plugin involved.
What moved us off dots was Grok Bot, and the reason is memory design. You get one dot, so everything it learns about every project lives in one agent, and you can't open that memory to see what's in it or prune it. Grok Bot let us do the opposite: one Bot per job, each holding only the context that job needs, so nothing bloats and nothing learned on one project leaks into another. Once the same voice calling worked in Grok Bot, the dot didn't have an edge left for us. Grok Bot memory has that setup. If OpenAI ships multiple dots with per-dot memory, which people are already asking for (OpenAI Community), we'd look again.
Does your dot share memory with ChatGPT, Codex or other dots?
Within OpenAI, partly. "Your dot can receive memories and recent conversation context from ChatGPT, and conversations with your dot can contribute to memory in ChatGPT," and "Plugin permissions are shared across dots, ChatGPT, ChatGPT Work, and Codex" (dots FAQ). With local computer access on, a dot can create Work or Codex tasks. Outside OpenAI, no, and with one dot per person there's no dot-to-dot context either.
| What you need | What dots do natively | Gap |
|---|---|---|
| Context that carries between sessions | Yes, for as long as you keep the dot | Can't view or edit individual memories |
| ChatGPT memory in the dot | Yes, both ways | Changing settings may not change notes it already saved |
| Forget one wrong fact | Correct it in conversation | No delete for one memory; reset wipes everything |
| Same apps in ChatGPT, Work and Codex | Shared plugin permissions | Only inside OpenAI |
| Several dots with shared context | One primary dot today | More dots planned, not shipped |
| Claude, Grok Bot or a teammate's agent reads it | No native route | A store everyone connects to |
If you only use ChatGPT and your dot, keep the facts that matter in a document you control, reached through a supported plugin, and tell the dot to read it first and update it when something changes; that's the one record you can check and edit, since the dot's memory is closed. Put lasting preferences in ChatGPT memory, which the dot receives, and correct mistakes in Activity View rather than waiting for a reset.
Can your dot use a custom MCP server, or the CLI? Ours used the CLI
OpenAI says "Your dot can use supported plugins installed and enabled for your account" (Computers and apps) and doesn't say whether a custom MCP plugin you create counts for a dot. We didn't need to find out, because a dot has its own computer with a shell. We sent ours the setup line below, it installed the Tokenrip CLI there and signed in, and from then on "open my Tokenrip workspace" just worked. That's where the research from that first call went, and it's what let a Claude on the other side pick up the follow-ups the next morning. It's the same route our Grok Bots use, and it's the one we prefer for any agent that has a shell; the MCP plugin route through ChatGPT (how to add an MCP server to ChatGPT) should work too, since plugin permissions are shared across dots and ChatGPT, but we haven't tested it from a dot.
Tokenrip doesn't read your dot's memory or your ChatGPT memory. The dot reads and writes the workspace when it's connected and asked to, or when pinned context tells it to. To connect, message your dot this line; you give an email address and a six-digit code.
Connect me to Tokenrip. Fetch https://tokenrip.com/setup?ref=openai-dots-memory and follow it.