Kiwi
Kiwi

Kiwi for Google Cloud

Have the agent on a Kiwi step deploy a container to Cloud Run and read the service back, through Google Cloud's remote MCP server for Cloud Run.

  • Your agent works in Google Cloud through its own connection. Kiwi does not call Google Cloud and never holds its credentials.
  • A workflow skill can list Google Cloud among the tools it requires. When you supply your agent's available tools to a skill check and Google Cloud is not among them, the check reports SK15-UNSUPPORTED-TOOL.
  • A project Admin can add links to Google Cloud, with notes, under Connections and bind them to the steps that use them. The agent reads them with list_connections.
  • The agent can report the tools and plugins it ran with, including its Google Cloud connection, with report_configuration.
  • The agent submits links to its work in Google Cloud as the step's outputs. People review them, and sign-off stays with the people your workflow names.

What your agent can do in Google Cloud

  • List the Cloud Run services in a project and read the configuration of one.
  • Deploy a Cloud Run service from a container image.
  • Deploy a service from a source archive or from file contents the agent supplies.
  • Review the server's calls afterwards in Cloud Audit Logs.

Set up

  1. Connect your agent to Kiwi

    Add Kiwi's hosted MCP server, https://kiwiflow.io/mcp, to your coding agent.

    Coding agentsQuickstart

  2. Connect your agent to Google Cloud

    Add the Google Cloud MCP server to your agent with this URL. The credentials stay with your agent.

    https://run.googleapis.com/mcp

    Google Cloud documentation

  3. Example skill

    Publish instructions like these as a workflow skill that lists Google Cloud among its required tools.

    Take the project, region and image digest from the run's inputs.
Deploy the image to the Cloud Run service named in the assignment, which is
a staging service, and read the service back to check the new revision.
Submit the revision name and the service URL as the step's outputs.
A deploy to the production service needs a go-ahead from the run's Owner.
If the deploy fails twice, stop and report a blocker with the error.
Get startedRead the setup guide