Add an AWS CloudWatch trace connection

  • Release version: Zurich
  • Updated June 30, 2026
  • 2 minutes to read
  • Monitor AI agents built with AWS Bedrock AgentCore by adding an AWS CloudWatch trace connection. AI Control Tower collects trace data through your AWS credentials and a MID Server, without requiring SDK instrumentation.

    Before you begin

    Confirm the following:

    • Your AI agents are built using AWS Bedrock AgentCore.
    • Tracing is enabled for each AWS Bedrock AgentCore agent you want to monitor.
    • An active MID Server is installed and configured in your ServiceNow instance. See MID Server installation.
    • An AWS access key ID and secret access key with permission to read trace data from CloudWatch is available.
      • This connector supports access key authentication only. Connections using an assumed role or external ID aren't currently supported.
      • The access key must be created and its IAM policy configured in AWS. For details, see the Work Instruction | How to Create AWS Credentials on a Glide Instance [KB3143433] article in Now Support.
      • After the access key ID and secret access key are created, work with your instance administrator to store them as a new AWS credential record in All > Connections & Credentials > Credentials.
    • You know the AWS region where your AI agents are deployed, such as us-east-2.

    Role required: sn_ai_governance.ai_steward

    Procedure

    1. Navigate to All > AI Control Tower > Home > Settings > Integrations > Traces.
    2. On the Available sub-tab, select AWS CloudWatch.
    3. On the form, fill in the fields.
      Table 1. Add AWS CloudWatch trace connection form
      Field Description
      Name Descriptive name for this trace connection. Use a name that distinguishes this connection from others you create, such as one that identifies the account, project, or environment.
      Collection frequency (minutes) Interval, in minutes, at which the MID Server polls for new trace data. The default is 30. Set a lower value to return results sooner or set a higher value to reduce overhead for lower-volume systems.
      Region AWS region where your AI agents are deployed, such as us-east-2.
      Credential AWS credential that authenticates AI Control Tower to CloudWatch. The credential must have permission to read trace data.
      MID server MID Server that runs trace collection. The MID Server must be active and validated. Select Go to Mid server installation to install or configure one.
      Log group source AWS CloudWatch log group that this connection collects trace data from. Select Agent Core Log Group to collect from the log group used by agents built with AWS Bedrock AgentCore. Select Other Log Groups to collect from a different log group. Select both to create a separate connection for each.
      Active Option to begin collecting traces when you save the connection. Clear it to save the connection without starting collection. You can activate the connection later from its record.
    4. Select Save.

    Result

    One or more trace connections appear on the Established sub-tab. If you selected both log group options, two connections appear, distinguished by a log group suffix on the connection name. If a connection is active, AI Control Tower begins collecting trace data after the first polling interval.

    What to do next

    Choose which metrics to include in evaluation scoring. See Activate evaluation scoring for external AI systems.