Skip to content
DocumentationQuickstartIntegrationsReferenceTroubleshooting
DEEN

Integration

ChatGPT

One address as a custom connector, one authorisation through your knowmind account — from then on ChatGPT looks things up in your memory while it answers.

Audience
Anyone using ChatGPT in the browser who may add custom connectors. On the knowmind side the free Private plan is enough.

Prerequisites

  • A ChatGPT account where custom connectors are available
  • In managed workspaces, approval by the administrator
  • An email address for the knowmind account
Note

Which ChatGPT plans allow custom connectors is set by OpenAI

Whether your plan offers custom connectors, and where the entry point sits, is OpenAI’s decision. The current state is in the OpenAI help centre. None of that changes the knowmind side: address and sign-in stay the same.

Steps

  1. 1

    Create your knowmind account

    Enter your email address at knowmind.de. You receive a sign-in link by mail and use it to sign in. There is no password.

    Then create your workspace: name, short name and plan. The Private plan costs 0 € and is enough for this guide; the pricing page lists the others.

    Outcome: The welcome screen is open, offering “start by writing” and “connect my AI tool”. Three starter cards are already in your workspace.

  2. 2

    Add the connector

    Create a custom connector in ChatGPT; the way there runs through the settings. Use this address:

    text
    https://knowmind.de/api/mcp/v1

    You sign in with your knowmind account. No access key goes in.

    Outcome: knowmind appears in your connector list.

  3. 3

    Authorise access

    On connecting, ChatGPT redirects you to the knowmind sign-in. You sign in with your sign-in link and confirm the permissions — reading only, or reading and writing. Then it returns to ChatGPT, which fetches the tool list.

    Outcome: ChatGPT shows the knowmind tools, and the Cockpit lists ChatGPT under connected applications.

  4. 4

    First recall

    Open a new chat, select the connector and ask: “Search knowmind for my open items.” ChatGPT calls knowmind_recall_memories.

    In a fresh workspace the three starter cards come back, marked as a starting point rather than a solid hit. Save something first: “Save to knowmind: the monthly report goes out on the third working day.” The next recall finds it straight away.

    Outcome: ChatGPT answers from your content and shows the tool call with its parameters and result.

Verify the result

The integration works when all three hold:

  • knowmind is in the connector list and switched on.
  • The knowmind tools can be selected in a new chat.
  • The Cockpit shows ChatGPT under connected applications and the recall in the timeline.

Troubleshooting

Error messageCauseResolution
There is no section for custom connectorsThe plan does not offer them, or the workspace administration has blocked them.The current state per plan is in the OpenAI help centre; in a managed workspace the administrator enables them. Without custom connectors, Cursor and the command line reach the same memory.
The connection reports the endpoint as unreachableA typo in the address.Check the address exactly: https://knowmind.de/api/mcp/v1. Availability shows at /api/health.
ChatGPT sees knowmind but does not use itThe model does not recognise on its own that a recall would help.Ask for it explicitly: “Search knowmind for …”. As a lasting fix, put a memory pack into your ChatGPT custom instructions (German page: Memory-Pack).
Calls return 401The authorisation was revoked or no longer applies.Disconnect it under connected applications and connect again in ChatGPT.
What we tested
Tested on 19 October 2026 against knowmind.de: account, workspace, client registration, consent, authorisation code exchange and first recall all work — with the real ChatGPT redirect. The wording inside the ChatGPT dialog and the plan limits are OpenAI’s; we did not verify those.

Related