Cloud Trace Connectors
Cloud Trace Connectors pull agent activity from AWS Bedrock and Google Vertex AI into AgenticAnts — so managed-platform agents show up as traces alongside your SDK-instrumented apps, without changing application code.
Enterprise feature. Connect from Settings → Connectors. You'll provide read access to your cloud platform and choose how traces are collected.
Why use a connector
If your agents run on Bedrock or Vertex, much of their telemetry already exists in your cloud account (logs, traces, exports). A connector reads that telemetry and turns it into platform traces — giving you observability, cost, and governance over managed-platform agents you didn't instrument by hand.
AWS Bedrock & GCP Vertex
Credentials: an AWS access key + secret with read access to the relevant services.
Collection methods:
| Method | Source |
|---|---|
| CloudWatch Logs | Bedrock invocation logs in CloudWatch |
| S3 Export | Invocation logs delivered to an S3 bucket |
| AgentCore (OTel) | OpenTelemetry from Bedrock AgentCore |
| Invoke / Flow APIs | Agent and flow execution via Bedrock APIs |
The connector can discover available Bedrock agents/flows so you can map them to a project.
Connecting
Provide credentials
In Settings → Connectors, add a Bedrock or Vertex connector with read-only cloud credentials. The connector validates that it has the permissions the chosen method needs.
Choose a collection method
Pick the method that matches how your telemetry is already emitted (e.g. CloudWatch for Bedrock, Cloud Trace for Vertex).
Map and ingest
Map discovered agents/flows to a project. Traces begin flowing in and appear in Observability.
Connectors use read-only access to your cloud account. Grant only the permissions the selected collection method requires — the connector tells you which are missing.
Next steps
- Observability — view the ingested traces.
- FinOps — cost and latency for managed-platform agents.
- AI Governance — govern Bedrock/Vertex agents alongside the rest.