This article provides a recipe for using the Ardoq Import Builder to automatically discover Agents in Google’s enterprise agent environment - Vertex AI. You can use the same pattern for doing discovery in other Agent runtimes such as Microsoft Foundry and Amazon Bedrock.
This approach speeds up the use of AI Lens for managing and governing the AI-enabled applications and agents in your organization.
Setting Up the Connection
Pre-requisites in Vertex AI
You need to have authorization permission to access the list of Agents in your Vertex AI projects. These are referred to as reasoning-engines . Use the gcloud CLI utility to get the access token for your projects
gcloud auth print-access-token
Test your access works by either accessing the reasoning-engines API directly or using the gcloud utility
curl -X GET \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \ "https://<apilocation>aiplatform.googleapis.com/v1/<parent>/reasoningEngines"
<apilocation>aiplatform.googleapis.com is your geo-location specific deployment for Vertex AI.
If your access is working then you can create the connection
Ardoq Connection
In Ardoq, open up the Import Builder, go to the Connections tab and create a new connection. Next, configure the following setting for a basic set up:
Set the Base URL to
<apilocation>aiplatform.googleapis.comSet the Authentication method to Bearer token
Set the token value to the access token:
gcloud auth print-access-tokenSet the endpoint to
/v1/<parent>/reasoningEnginesRun the request to verify your connection and click Save connection
The configuration should look like the following:
Discovering and Importing Agents with the Connection
Create an integration connection, assign the above-created connection, and select the Agent list for the data mapping
Import Components and then assign to an existing workspace or create a new workspace for Agents
You can then review your imported agents and create the references from AI Agents Technical Capability to make them managed and governed in AI Lens
Once your agents are imported, you can place them in context by linking each one to the wider AI technology landscape in Ardoq, connecting AI Agents to the underlying AI capabilities they rely on (e.g., LLMs, multimodal or diffusion models, and applied AI platforms) as well as the broader technical capabilities they enable.
This makes AI Lens more than an inventory: it becomes a map of how agents contribute to business outcomes, where shared dependencies and governance risks sit, and how your agent portfolio fits into the organization’s overall AI maturity and architecture (as illustrated in the capability view below).
