AI Studio Connectors
AI Studio connectors pull agent telemetry from your cloud account and turn it into ANTS Platform traces. They also list every agent in the cloud account, so agents appear in the inventory before they first run. Use them for agents that run on a managed platform and are not instrumented with an SDK.
| Platform | What it reads | What it lists | Guide |
|---|---|---|---|
| AWS Bedrock | Bedrock Agent trace events from CloudWatch Logs, or Bedrock Flow executions | Every Bedrock agent and flow in the account and region | AWS Bedrock |
| Google Vertex AI | Vertex AI Agent Engine (reasoning engine) logs from Cloud Logging and spans from Cloud Trace | Every reasoning engine in the Google Cloud project and region | Google Vertex AI |
For Microsoft 365 Copilot and Copilot Studio, see Microsoft 365 Copilot.
Where to configure
Go to Project Settings > Integrations and scroll to AI Studio connectors. The same page is linked from Project Settings > API Keys under External agent integrations.
You need the Owner role, or a custom role with write access to AI Studio Connectors under Integrations. Read access shows the list and sync logs. See Members & Roles.
Deployment requirement. Each platform has its own connector service.
On self-hosted deployments they are off by default. Set
bedrockConnector.enabled and vertexConnector.enabled to true in the
Helm values to run them. Without the service, connectors can be created but
never sync.
How connectors work
- Create. You enter cloud credentials and the agent or flow to read. ANTS Platform checks the credentials by fetching the agent or flow name, which becomes the connector's display name. It also creates a dedicated project API key that the connector uses to send traces into this project.
- Sync. The connector service polls on the Sync Frequency you choose (every 5, 15, 30 or 60 minutes; default 15). Each sync reads events after the last checkpoint and sends them as traces with generation and span observations.
- Checkpoint. After a successful sync, the checkpoint moves forward, so events are not imported twice.
What the first sync covers depends on the platform. See the platform guides.
Agent inventory
Traces show an agent only once it runs. To cover agents that exist but have not run yet, or run rarely, each sync also lists every agent in the cloud account with the connector's credentials:
- AWS Bedrock: every agent and flow in the account and region
(
bedrock:ListAgents,bedrock:ListFlows). - Google Vertex AI: every reasoning engine in the project and region
(
aiplatform.reasoningEngines.list, included inroles/aiplatform.viewer).
Listed agents are added to the AI Agent Center. Those never seen in traces are marked Not seen running, until their first traces are imported.
Manage connectors
The AI Studio connectors list shows each connector's platform, method, Last sync, Status and an Active switch. From the row you can:
- turn Active off to pause syncing, and on to resume,
- Edit connector to change the Display Name, Sync Frequency or credentials,
- View sync logs to see when each sync started, its status, the traces pulled, ingested and skipped, and any error,
- delete the connector.
Credentials are encrypted at rest and never shown again.