Import an existing agent
Discover and import AI agents already deployed on Google Agent Runtime or AWS Agent Core into Harness as Agent Services.
If your team already has AI agents running on Google Agent Runtime or AWS Agent Core, you do not need to redeploy them to bring them under Harness governance. The Import an Agent flow discovers agents already deployed on either runtime, registers them as Agent Services in Harness, and then lets you manage, promote, and roll back those agents through Harness pipelines going forward.
Before you begin
Harness account with agent deployments enabled: When the feature is enabled, Agent Services appears in the left navigation of the Continuous Delivery & GitOps module. Contact your account administrator or Harness Support to enable
CDS_AGENT_RUNTIME_DEPLOYMENT.A connector with access to your cloud platform: You need a connector already configured in Harness Settings. For AWS Agent Core, you need an AWS connector. For Google Agent Runtime, you need a GCP connector. Go to AWS connector settings reference or Google Cloud connector settings reference to create one.
Agents already deployed on the target runtime: The agents must already exist on Google Agent Runtime or AWS Agent Core. Harness reads their configuration from the runtime at import time.
Import an agent
In the Continuous Delivery & GitOps module, select Agent Services in the left navigation.
Select Import an Agent in the top right of the listing page. The Import an Agent panel opens.
Under Platform, select the runtime where your agent is deployed:
AWS Agent Core — Bedrock AgentCore runtime
Google Agent Runtime — Vertex Agent Engine
Under Connector, select the connector for your cloud account. To add or manage connectors, go to Settings. The panel shows a note that connectors are configured in Settings.
Under Scope, provide the location details for your deployed agent. The fields differ by platform:
AWS Agent Core: Select the Region where the agent is deployed, for example
US East (N. Virginia).Google Agent Runtime: Select the Project and then the Location where the agent is deployed.
In the Agent dropdown, select the agent you want to import. The dropdown lists all agents discovered in the selected scope, showing each agent's name and its platform identifier. If you do not see your agent, verify that the connector has the correct permissions for that scope.
In Agent name, review or edit the name that Harness will use for the imported Agent Service. Harness pre-fills this from the agent's name on the runtime. You can also edit the Id before importing. If you change the name from the discovered default, Harness records the mapping so that reconciliation during future redeployments correctly identifies the agent on the runtime.
Select Import an agent.
Harness creates the Agent Service and opens its Summary tab. The About section shows the agent description and artifact (the container image URI). The Deployments section is empty until you run a Harness-managed deployment for this agent.
After import
Once an agent is imported, it appears in the Agent Services listing like any agent you registered from scratch. From the Configuration tab you can:
Edit the image URI, execution role ARN, config variables, or AgentConfig manifest. Any changes take effect on the next deployment.
Add or update the manifest. Go to Add an AgentConfig manifest for details on the manifest option for AWS Agent Core.
To manage the agent through a pipeline, create a pipeline with a Deploy stage that references the imported Agent Service, then set up an environment with an infrastructure definition that targets the same runtime. Go to Google Agent Runtime: Canary deployment or AWS Agent Core: Canary deployment to set up the pipeline.
Importing an agent does not affect the running agent. Harness reads the configuration at import time, but the agent continues to run on the runtime until you trigger a Harness pipeline that deploys to the same infrastructure.
Next steps
After importing an agent, configure an environment and pipeline to manage its deployments through Harness.
Last updated
Was this helpful?