Add support for deployment on Agent Engine

[ipeychev]

Most probably this is in the roadmap already, but GitHub doesn't return anything on this topic, so I will open it. Please, feel free to point me to any duplication and I'll close this one.

Problem
The only option for deploying an Agent seems to be on Cloud Run.

npx adk deploy --help          
Usage: ADK CLI deploy [options] [command]

Deploy agent

Commands:
  cloud_run [options] [agents_dir]
  help [command]                    display help for command

Describe the solution you'd like
We need to add support for deployment on Agent Engine

[Varun-S10]

Hi @ipeychev, We would appreciate more information about your specific requirements for this feature. Any details you can share would be very helpful.

[ipeychev]

Hey @Varun-S10 , thank you for pinging me. Here it is:

1. Registration in Gemini Enterprise (ex-Agentspace). We need to register our TypeScript agents in the Gemini Enterprise console so our employees can use them.

• Currently, this registration requires the agent to be a "Reasoning Engine" resource.
• In Python, this is done natively using the command adk deploy agent_engine or the SDK method .create().

We need the TypeScript ADK to support this managed deployment flow so we get a first-class Google Cloud resource ID for Gemini Enterprise.

2. Managed Session Service (VertexAiSessionService)

adk-js is missing a native VertexAiSessionService. Using InMemorySessionService is not viable for production - conversation history is lost if the Cloud Run instance restarts or scales down. The session.state needs to be stored in Google’s managed infrastructure (as it is in Python).

3. The ability to revert a session to a previous state or event ID in the history.

4. Native IAM integration so the agent uses the Cloud Service Account identity for tool calls and session management without custom code.

5. CLI Deployment Support
   As shown in the help output, the current adk-js CLI only supports Cloud Run, which does not automatically register the agent as a Reasoning Engine:

npx adk deploy --help
# ...
Commands:
  cloud_run [options] [agents_dir]

We need a reasoning_engine (Agent Engine) command to match the Python CLI. This should handle the containerization and registration in Vertex AI automatically.

In general, we want to write our logic in TypeScript but keep the "managed" experience Google provides for Python agents.

6. The TypeScript ADK should natively support Agent Identity.

When deployed to Agent Engine, the agent should automatically assume the identity of its associated Service Account for all tool calls (e.g., querying BigQuery or Vertex Search) without the developer manually initializing GoogleAuth within every tool.

Hope that helps!

[kalenkevich]

Hello! Agent Engine support for JavaScript/TypeScript is currently unavailable. This is a top priority for our team, and we plan to implement deployment capabilities once the underlying support is released. Thank you for your patience!

[dnlrbz]

The documentation on how to deploy an adk-js agent to google cloud would also be very helpful 🙏 just like it exists for Python

[kalenkevich]

@dnlrbz Hello, sure we will update that soon. Thanks for pointing that out.

Meanwhile please, try this:

Create a folder with the next structure:

./your_project
  -- package.json
  -- agent.{js|ts}

In the package.json ensure you have @google/adk as a dependency and all other required dependencies for your agent (e.g. zod).

Deploy command is pretty much the same as for other ADKs: npx @google/adk deploy cloud_run ./your_project ...flags where flags are:
--project TEXT: (Required) Your Google Cloud project ID (e.g., $GOOGLE_CLOUD_PROJECT).
--region TEXT: (Required) The Google Cloud location for deployment (e.g., $GOOGLE_CLOUD_LOCATION, us-central1).
--service_name TEXT: (Optional) The name for the Cloud Run service (e.g., $SERVICE_NAME). Defaults to adk-default-service-name.
--port INTEGER: (Optional) The port number the ADK API server will listen on within the container. Defaults to 8000.
--with_ui: (Optional) If included, deploys the ADK dev UI alongside the agent API server. By default, only the API server is deployed.
--temp_folder TEXT: (Optional) Specifies a directory for storing intermediate files generated during the deployment process. Defaults to a timestamped folder in the system's temporary directory. (Note: This option is generally not needed unless troubleshooting issues).

Note 1: ADK TS CLI will compile and bundle the your agent.ts to javascript to achieve best compatibility with older nodejs versions that does not support native typescript execution.

Note 2: --session_service_uri is not really supported yet but will be very soon as Database Service is WIP: #129 and will be included in the one of the upcoming releases.

Note 3: --agent_engine_id is not supported yet due to #85 (comment)

Hope that will unblock you for now.
Thanks!