Skip to content
Development Working integration

The IntraGPT API portal: let your own systems talk to the AI environment

The API portal gives your own software access to the AI environment. You create a token in the dashboard, hand it to your intranet, portal, app or script, and from there you ask a question or read a conversation back. Everything arriving through such a token stays inside your workspace, with the same knowledge and the same boundaries as in the chat.

In most organisations the AI environment stands on its own: there is a place you go to in order to ask something. That works fine right up to the moment the question arises somewhere else. In the customer portal, in the service desk application, in a nightly script walking through hundreds of documents. At that point you do not want an employee opening a second window and retyping answers. That is what the API portal is for. It turns the AI environment into a service your own software can call. You ask something from your own system and the answer comes back into that system. In practice organisations use that in two ways. The first is letting a piece of existing software hold a conversation: an employee portal that forwards a question and shows the answer in its own house style. The second is work in bulk: a script that asks a series of questions and stores the answers, or an overnight run with nobody at the controls. What matters is that none of the care is lost. A token belongs to a single workspace, with the same knowledge sources and the same rights. There is no back door around the boundaries.

What the portal gives you

A token per integration

You create a token per connecting system, with a recognisable name. That way you know later which token belongs to which system and you can revoke one without the other integrations dropping out.

Hold a conversation from your own software

Your own system asks a question and receives the answer. Follow-up questions belong to the same conversation, so context is preserved, exactly as in the chat.

Read a conversation back

An earlier conversation can be retrieved with its messages, including who said what and when. Useful when you want to show the exchange in your own application or archive it.

Manage your own assistants

The assistants you configured in IntraGPT can also be listed and managed programmatically. A management screen in your own environment can therefore show the same as the dashboard.

Documentation inside the portal

The portal holds a page with an example per part of what you send and what comes back. Your developer needs no separate documentation package to get started.

What organisations use this for

Customer service

Situation
The service desk works in its own application and wants a proposed reply right there, not in a separate window.
What the agent does
The application forwards the customer question and shows the proposal in the screen the agent is already in.
Result
Less switching between systems, and an answer grounded in the organisation's own knowledge.

Intranet

Situation
Employees prefer asking their questions on the intranet rather than in a separate environment.
What the agent does
The intranet forwards the question and shows the answer in its own layout, with the source reference attached.
Result
The AI environment gets used without anyone having to learn a new place to go.

Information management

Situation
A large stack of documents has to receive the same treatment, a summary per item for instance.
What the agent does
A script asks the questions one after another and stores the answers in the organisation's own system.
Result
Work that would otherwise take weeks runs overnight, with a checkable result per document.

Software development

Situation
An in-house application needs to show which assistants are available to a user.
What the agent does
The application retrieves the assistants and shows them in its own management screen.
Result
One place where administrators work, instead of two screens drifting apart.

From token to integration

  1. 1

    Switching the portal on

    The API portal is enabled per organisation. With it off the whole section does not exist and no token can exist either. Turning it on is a decision your administrator makes.

  2. 2

    Creating and copying a token

    In the portal you give a token a name and copy it. From that moment the system you put it in can ask questions. Renaming and revoking are possible at any time, without the rest suffering.

  3. 3

    Deciding what the integration sees

    A token belongs to one workspace and therefore to the knowledge sources and assistants of that workspace. What is not in there does not come out of it. There is no route by which a token reaches another organisation's data.

  4. 4

    Watching what comes in

    Conversations started through an integration simply sit in the environment, like conversations from the chat. An administrator can therefore see afterwards what was asked and what the answer was.

Connected securely

A door you open and close yourself

A token from the API portal is a key that lives in software rather than in an employee's head. That is why it matters that you can create several and revoke just one.

Where the data lives: Nederland of de EU

One token per system

By giving every connecting system its own token, a revocation stays small. If a token goes missing somewhere, you revoke that one and the rest keeps working.

Bolted to one workspace

What arrives through a token stays within the workspace that token belongs to. The boundaries that apply in the chat apply here too, including the knowledge sources hanging off it.

Switchable at organisation level

If an organisation wants no programmatic access at all, the portal simply stays off. It is not a setting that can be on by accident.

All on our own server

What arrives through an integration is processed on our own server in the Netherlands as well. It does not go to an American model vendor and is not used for training anywhere.

Questions about the API portal

Does IntraGPT work with our own software?

Yes, that is what the API portal is for. You create a token and your own intranet, portal, app or script can then ask a question, continue a conversation, read one back and list your own assistants.

Can I use this to connect IntraGPT to ChatGPT?

That is not what it is meant for. The portal gives your own systems access to your own AI environment, with your knowledge and your boundaries. Giving a public chatbot access to that environment from the outside is exactly the scenario we want to avoid.

How do I revoke a token?

The portal lists the tokens by name. You remove the one you no longer want and from that moment it stops working. Because you create a token per system, that does not affect the other integrations.

Can a token reach another organisation's data?

No. A token belongs to one workspace and therefore to the knowledge sources and assistants of that workspace. There is no route by which it can look past that boundary.

Do we need a developer for this?

Not for creating a token, an administrator does that in the dashboard. For the other side you do: somebody has to build it into your own software. The portal holds documentation with examples so that does not turn into a search.

Do conversations started through an integration show up in the environment?

Yes. They sit there like conversations from the chat, with the question and the answer. That is pleasant while setting things up, because you see exactly what your own system is asking.

What does using the API portal cost?

The portal is part of the track we go through together, not a separate line item. What drives the work is how many systems you want to connect and what those systems need to do. We lay that out in a thirty minute conversation.

How long does setup take?

Switching it on and creating a token takes minutes. The lead time sits on your side: somebody has to build the call into your own software and test it. For a first working connection that is usually half a day.

For which sectors

Other connectors

See all connectors

Curious what this connector would deliver for you?

Book a free thirty minute AI call. We look at your systems, the permissions around them and the first use case that saves time or money.

Book a free AI consultation