Setting MetronHR up in ChatGPT

In the ChatGPT settings create a connector of your own, enter app.metronhr.de/mcp as the address and choose OAuth as the authentication. Client ID and secret stay empty. Then sign in once in the browser, and ChatGPT answers questions about your own business's hours, absences and projects.

As of:15. September 2026

Setting it up in four steps

Management does the first two steps in MetronHR, and each person does the last two in their own ChatGPT.

  1. 1. Switch AI connections on for the business

    Company settings, Integrations section. The default is off. Beside it stands which roles may connect; an empty list means nobody.

  2. 2. Decide whether writing is allowed

    Its own switch with its own list of roles. Without it ChatGPT can read and nothing else. It can be withdrawn at any time and takes effect at once.

  3. 3. Create the connector in ChatGPT

    Open the connectors area in the settings, create one of your own, enterhttps://app.metronhr.de/mcp as the address and choose OAuth as the authentication. Client ID and secret stay empty. Where exactly that area sits and what it is called is something OpenAI changes from time to time.

  4. 4. Approve it in the browser

    ChatGPT opens the sign-in flow. Anyone not yet signed in to MetronHR signs in; the consent dialog then names in plain words the rights that remain under the business's rule.

What that is good for day to day

The assistant comes with two ready-made reports and can ask freely alongside them. The examples are the questions the templates exist for.

  • Check the month end for me.

    What happens behind it
    The report that ships with it fetches the period's time entries and absences and names what stands out: missing days, open requests, hours without a project.

  • Who built up overtime last quarter?

    What happens behind it
    The second report calculates per department against the working time model stored in MetronHR, not against an assumption.

  • Which holiday requests are open?

    What happens behind it
    A read tool on absences. The type goes along, the note on a sick report does not.

  • Book two hours to project X for me.

    What happens behind it
    Only with writing released. The call runs through the same rules as an entry in the product and is then in the log.

As of 15. September 2026. What a connector can actually fetch depends on the rights of the person connecting and on the business's release.

When it does not work

Three causes cover almost everything. The first sits with OpenAI, the other two in the company settings.

  • There is no entry for custom connectors

    Cause
    The ChatGPT plan does not carry them. That is OpenAI's decision, not MetronHR's.

  • The sign-in ends with a refusal

    Cause
    AI connections are off for the business, or the person's role is not on the list.

  • The assistant reads but changes nothing

    Cause
    Writing is a release of its own. It is missing, or it does not cover this role.

  • After a while nothing comes back

    Cause
    The connection was disconnected, or a release was withdrawn. Either takes effect with the next request.

As of 15. September 2026. Every tool call is in the business's log, with the tool and the client. That also shows whether a request arrived at all.

The address, the protocol revision and the permission model are the same for every assistant; they are set out on the page about the MCP server.

Questions about ChatGPT and MetronHR

What we are asked most often about it.

One in which custom connectors are available. Which those are is OpenAI's decision and changes there too; in the free plan they are not. That is the one part of the setup a business cannot solve inside MetronHR.

No, both fields stay empty. ChatGPT registers itself with the MetronHR authorisation server and fetches the addresses it needs from the metadata. All you enter is the server's address.

It sees what the person connecting may see, and of that only the released resources. The consent dialog names the rights in plain words before the connection is made. Health information, bank details and social insurance numbers never leave the product.

Only if the business has released writing, and only for the roles listed there. Without that second release the connector can read and nothing else. A write call then runs through the same rules as an entry in the product.

What the assistant fetches, it processes at its own provider. That decision comes before the connection and belongs in the data processing agreement with OpenAI. MetronHR hosts in Germany and passes nothing on of its own accord.

In MetronHR under company settings, Integrations section: disconnect that connection on its own. Access and refresh tokens become invalid at once, and the connector in ChatGPT loses access with its next request.

Question not answered here? Every step is explained in the help centre.

Ask ChatGPT about your own figures

The connector is ready in two minutes, and the release stays with the business.

See the prices

No credit card, cancel any time

Related

Where to go from here, and what waits for you there.