Pricing Question about Vertex AI Agent Engine compute and memory behavior

Hello,

I need clarification regarding Vertex AI Agent Engine pricing, specifically for compute and memory usage.

    • Is the compute and memory pricing charged continuously, even if the agent is not actively processing any requests?
    • Does the number of requests processed by the agent affect the compute and memory charges?
    • How can I configure or limit the CPU and memory resources allocated to an agent to control costs?
    • My agent is very small and idle most of the time β€” I want to avoid paying continuous hourly charges if it’s not used.
    • In terms of cost efficiency while idle, is it better to deploy an ADK agent on Cloud Run instead of using Agent Engine?

Thank you!

5 Likes

You can find direct answers to most of your questions on the following docs pages:

Vertex AI Agent Engine pricing

Deploy an agent on Vertex AI Agent Engine (customised resource control section)