Deployment & infraDatabases & storageAI modelsAnalytics & product

Google Cloud

Cloud APIs, gcloud CLI, IAM, BigQuery, GKE, and AI infrastructure.

Agents can use gcloud and APIs for cloud resource management when constrained by projects, IAM, and dry-run plans.

Agent access evidence

CLI

native

gcloud CLI is scriptable and includes bq, gsutil, and kubectl tooling.

Evidence

API

native

Google Cloud APIs cover platform services with client libraries.

Evidence

MCP

native

Google and Google Cloud provide remote MCP servers for supported Google products and cloud services.

Evidence

Browser

strong

Console is important for billing, IAM, logs, and audits.

Evidence

Account creation

strong

Self-serve account/project creation with billing setup.

Evidence

Pricing clarity

strong

Public per-service pricing and calculator, but cross-service cost can be complex.

Evidence

Docs quality

strong

Docs and reference pages are broad and service-specific.

Evidence

Sandbox

partial

Separate projects and IAM roles support safer experimentation.

Evidence

Readiness score

Strong

Strong: 78/100. Strongest in machine operability; weakest in production reliability.

Machine operability24/25

Agent safety17.4/25

Agent readability17.4/20

Auth and setup13/15

Production reliability6.2/15

Best for

Cloud Run, BigQuery, GKE, storage, analytics, and AI platform operations.

Pricing

Public per-service pricing and calculator, but cross-service cost can be complex.

Auth

Service accounts, OAuth, workload identity, IAM roles, and ADC.

Account creation

Self-serve account/project creation with billing setup.

Browser support

Console is important for billing, IAM, logs, and audits.

CLI

gcloud

API

https://cloud.googleapis.com

MCP

https://docs.cloud.google.com/mcp

Limitations

Key limitations: IAM, billing, production data, and infrastructure changes need human approval. Model behavior, costs, rate limits, data retention, and safety policies can change by model and account tier; use evals, spend caps, and data controls for sensitive workloads. MCP servers inherit the connected account's scopes and may send prompts or data to third-party services; configure read-only or narrow scopes when available. CLI access is powerful but can run destructive local or production commands; prefer dry runs, preview environments, and non-production credentials.