ChatGPT Dots and your knowledge base

What a dot is

A dot is an always-on agent in ChatGPT that can take a project from an idea to work you can use, according to OpenAI's getting started guide. It starts with context from ChatGPT memory and learns how you like to work. Dots are rolling out to Pro users aged 18 and older in markets excluding the European Economic Area, Switzerland, and the UK. They are also available to Business Premium users in supported ChatGPT regions, subject to workspace eligibility requirements. Starting November 18, dots become available to workspaces by default unless the admin opts out.

How a dot learns: memory and connected apps

Connected apps give a dot access to information it can use for your tasks (OpenAI). The dot can review that information proactively and form memories from it, even when you have not asked a specific question. It also receives memories from ChatGPT and can create its own memories, including from connected apps.

Control over those memories is limited. You currently cannot view, delete, or directly modify individual dot memories, including details that entered the dot's context from plugins (OpenAI's privacy FAQ). To delete a dot's own saved memories, you need to delete the dot (OpenAI).

Plugins and shared permissions

The Plugins screen in a dot uses shared ChatGPT settings, and available apps may differ (OpenAI). Plugin permissions are shared across dots, ChatGPT, ChatGPT Work, and Codex, so a dot can use your existing connections within the permissions you have granted (OpenAI's privacy FAQ).

One more limit to know: disconnecting an app does not delete information your dot has already obtained from it (OpenAI).

The dot's computer and local skills

Your dot has its own cloud computer. Access to your local computer is optional and starts turned off, and connecting a computer requires the ChatGPT desktop app on that machine (OpenAI). Local skills require a computer connected to your dot (OpenAI).

This matters for personal knowledge bases. A library that lives on your own machine is not reachable by default; local access is an opt-in extra, not the starting state.

Custom MCP servers in ChatGPT

ChatGPT does not connect to local MCP servers directly. It connects to remote MCP servers, and servers on a private network, on-premises, or on a developer machine need Secure MCP Tunnel (OpenAI Help). Adding a custom server means entering a public HTTPS MCP endpoint URL, or selecting a tunnel and its tunnel_id (OpenAI Help).

The supported MCP protocols for custom servers are SSE and streaming HTTP (OpenAI's developer docs). Stdio is not among them.

Plans and admin approval

Custom MCP server connections and full MCP support are available for ChatGPT Business and Enterprise/Edu customers on ChatGPT web (OpenAI Help). Pro users can connect custom MCP servers with read and fetch permissions only (OpenAI Help).

On Business workspaces, only admins and owners can create and publish plugins for custom MCP servers (OpenAI Help). After approval, ChatGPT uses a frozen snapshot of the server's tools, and later changes apply only after an admin reviews and publishes an update (OpenAI Help). Company knowledge supports custom MCP servers with fetch and search access (OpenAI Help).

Can KnowMan connect to a dot today?

KnowMan runs locally with npx -y @knowman/mcp and as a hosted server at https://api.knowman.io/mcp with OAuth sign-in or an API key; see the MCP server guide. ChatGPT's custom server support lists SSE and streaming HTTP only (OpenAI's developer docs), and the hosted KnowMan endpoint speaks streaming HTTP. The dot guide does not describe connecting custom MCP servers to dots; it describes a shared Plugins screen where available apps may differ (OpenAI).

Secure MCP Tunnel connects private-network servers to supported OpenAI products (OpenAI Help), and the dot pages do not say whether dots are among them. There is no documented path from KnowMan to a dot today.

Where KnowMan fits

KnowMan is a personal knowledge base. It saves articles and links, reads and tags them, answers questions with citations, and keeps memories such as claims, decisions, and playbooks. Its MCP server runs locally with npx -y @knowman/mcp and as a hosted server at https://api.knowman.io/mcp with OAuth sign-in or an API key, with documented configs for Claude Desktop, Claude Code, and Cursor; see Claude Desktop setup and Claude Code setup.

If dots are not available to you yet, other routes exist today. Give your AI context and how to give ChatGPT memory cover working with ChatGPT, and a knowledge base for AI agents covers agent setups that run local servers.

Use the hosted KnowMan server in ChatGPT

KnowMan's hosted MCP server runs at https://api.knowman.io/mcp over streaming HTTP, the transport ChatGPT lists for custom MCP servers. To add it, enter https://api.knowman.io/mcp as the server URL in ChatGPT's custom MCP server settings. You need a KnowMan account, which you create at https://app.knowman.io.

Sign-in uses OAuth: the client opens a KnowMan consent page where you approve two scopes, knowman.read for search, read, and ask, and knowman.write for saving links. Access tokens last one hour and refresh automatically. To disconnect later, delete the entry shown as "OAuth: <app name>" in your KnowMan API-keys list.

The other documented option is an API key sent as an Authorization: Bearer header, created in KnowMan settings; a key cannot create other keys. ChatGPT's connector flow uses OAuth, so the consent page is the sign-in path there. The client behind both the hosted and local modes is open source at knowman-mcp.