Connecting an AI app
Connect Claude, ChatGPT or Microsoft Copilot Studio to Vectis as yourself, understand the consent screen, and disconnect.
An administrator must turn the connector on first.
Get the server URL
You need the Vectis server URL. An administrator can copy it from Settings → AI Connector → Server URL.
Connect your AI app
- In Claude, go to Settings → Connectors → Add custom connector.
- Paste the Vectis server URL.
- Select Connect. You are taken to Vectis to sign in (or to your organisation's own sign-in, if you use single sign-on), then to the Vectis consent screen.
- Review and approve. See the consent screen below.
On Claude Team and Enterprise, an organisation owner first adds the connector for the whole organisation from Claude's admin settings.
Result: Claude lists the Vectis tools, and a question such as "which vehicles have a WoF due this month?" is answered from your Vectis data.
- In ChatGPT, go to Settings → Apps & Connectors and turn on Developer mode.
- Create or add a connector, paste the Vectis server URL, and choose OAuth as the authentication method.
- Sign in to Vectis and approve on the consent screen.
Result: answers are drawn from Vectis. ChatGPT's Deep research and company knowledge features can also search Vectis, and their answers link back into Vectis.
- Open your Copilot Studio agent and go to Tools → Add a tool → New tool → Model Context Protocol.
- Give it a name and description, and paste the Vectis server URL.
- Choose OAuth 2.0, then Dynamic discovery, and select Create.
- Create a connection and sign in to Vectis to approve it.
- Add the tool to your agent, then publish the agent to Microsoft 365 Copilot or Teams.
Copilot Studio can't ask you to confirm a change itself, so every proposed change comes back as a link to approve in Vectis.
A Power Platform data loss prevention (DLP) policy can block custom MCP connectors. If the tool doesn't appear as an option, ask your Microsoft 365 administrator to check your Power Platform data policies.
The consent screen
After you sign in, Vectis shows a consent screen before anything connects:
- The app: its name and logo, with a Verified app badge, or an Unverified app warning if Vectis hasn't checked it. Only continue with an unverified app if you trust it.
- The organisation you are connecting it to.
- What it will be able to do: a list of permissions in plain language, such as "View your fleet" or "Propose booking changes". Untick anything you don't want to grant. Change permissions are greyed out if your organisation hasn't turned on Allow AI apps to propose changes.
- A data notice: the app receives the Vectis data it asks for while you use it, and your AI provider's own terms apply to that data. If your organisation keeps its data within Australian residency, the notice also says that some data will leave that boundary.
Select Connect to approve, or Cancel to back out. Nothing is connected until you select Connect.
Asking questions
Once connected, ask in normal language. The AI app works out which Vectis tool to use. For example:
- "Which vehicles have a WoF due this month?"
- "What's the current location of KXR412?"
- "List my pending proposals."
- "Who's driving vehicle ABC123 today?"
- "What's our current RUC status for the fleet?"
Answers only include what your Vectis role and pool assignments let you see. If you are limited to certain pools, you only get results for those pools, and looking up a record outside them is refused.
Disconnecting
Open your Profile and find Connected AI apps. Each app shows its name, a Verified or Unverified badge, when it was connected and last used, and the permissions it has. Select Disconnect and confirm.
Go to Profile → Connected AI apps. Tap the delete icon next to an app and confirm Disconnect.
The AI app's access stops at once: its next call to Vectis fails, and it will prompt you to reconnect if you want to use it again. To reconnect, follow the same steps as the first time, from the AI app itself.