Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsSome links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Google Cloud MCP Server is not one universal server. It is a portfolio of Google-managed, product-specific remote MCP endpoints. An AI client connects to an HTTP endpoint such as https://bigquery.googleapis.com/mcp or https://run.googleapis.com/mcp, then calls only the tools exposed by that Google Cloud service and permitted by IAM.
The practical setup is: choose the service, enable its API, configure a supported Google identity, grant roles/mcp.toolUser plus the service permissions required by each tool, and verify the product-specific MCP reference before allowing an agent to act on production resources.
What is the Google Cloud MCP server?
Google Cloud uses the term for managed remote MCP servers hosted on Google infrastructure. They provide HTTP endpoints through which an AI application communicates with a service-specific MCP server. The endpoint is not a general Google Cloud shell: BigQuery, Cloud Storage, IAM, Cloud Run and other products expose different tools, toolsets and operations.
Recommended Free Tools
Google’s catalogue changes frequently. The catalogue updated September 28, 2026 lists examples including:
#1 Best Overall
| Service | Example MCP endpoint |
|---|---|
| BigQuery | https://bigquery.googleapis.com/mcp |
| Cloud Run | https://run.googleapis.com/mcp |
| Cloud Storage | https://storage.googleapis.com/storage/mcp |
| Cloud SQL | https://sqladmin.googleapis.com/mcp |
| Cloud Logging | https://logging.googleapis.com/mcp |
| Cloud Monitoring | https://monitoring.googleapis.com/mcp |
| Compute Engine | https://compute.googleapis.com/mcp |
| IAM | https://iam.googleapis.com/mcp |
These are examples, not a complete inventory. Some catalogue entries are regional or marked Preview, so check Google’s live supported-products catalogue and the selected product’s MCP reference immediately before deployment.
Which Google Cloud services support MCP?
Supported entries include BigQuery, Bigtable, Cloud Run, Cloud Storage, Cloud SQL, Cloud Logging, Cloud Monitoring, Compute Engine, IAM, GKE, Pub/Sub and Spanner, among others. Availability, endpoint geography, toolsets and preview status differ. Select the service first; do not assume that an endpoint offers generic resource management or the same read/write behavior as another endpoint.
How do I connect an AI agent to Google Cloud with MCP?
- Select the endpoint and project. Identify the exact product operation the agent needs and record its MCP endpoint. Enable that product’s API in the Google Cloud project. Google’s introductory Cloud Logging exercise requires a project with billing enabled, the Logging API enabled, Application Default Credentials (ADC), and the relevant IAM grants.
- Check client authentication support. Google documents ADC, OAuth 2.0 client ID and secret, and an HTTP authorization header containing a bearer token. The AI client must support the method you choose. Google’s remote servers do not support Dynamic Client Registration or OAuth Client ID Metadata Documents.
- Create a narrowly scoped identity. You can use a user, workload/application identity or agent identity. Service-account impersonation is an option when your operating model requires it. A separate production agent identity makes review and revocation easier than sharing a personal account.
- Grant MCP permission. Give the identity
roles/mcp.toolUser, or a custom/predefined role containingmcp.tools.call. Google describes this permission as required to make MCP tool calls. - Grant service permissions. Add only the IAM roles required by the selected tool and target resource. MCP permission does not grant BigQuery, Storage, IAM or Compute access by itself.
- Configure the client and test a harmless operation. Use the endpoint, identity and project settings in your MCP-compatible AI application. Start with a read operation, inspect the returned tool list and confirm that audit records identify the intended principal.
How do I authenticate to a Google Cloud MCP server?
Application Default Credentials
ADC is convenient for Google-aware local development and supported hosted workloads. Authenticate the account used by the client, ensure its project context is correct, and verify that the resulting credentials are the same identity to which you granted MCP and service roles. ADC availability is a client and runtime concern; an MCP application that cannot read ADC needs another supported method.
OAuth 2.0 client credentials
Some clients can be configured with an OAuth 2.0 client ID and secret. Store secrets in the client’s secret-management facility, not in prompts, source control or shared configuration. Confirm the OAuth scopes and redirect or consent requirements expected by that client.
Bearer-token authorization
A client may send an HTTP Authorization: Bearer ... header. Tokens must represent an identity that has both mcp.tools.call and the underlying product permissions. A valid token alone does not authorize an operation.
Google’s current overview describes a stateless request model for MCP version 2026-07-28. The IAM guidance says required information is carried in HTTP headers or _meta, rather than relying on the older initialize handshake and session ID pattern. Protocol documentation is version-sensitive; confirm the behavior supported by your client and endpoint.
What IAM role do I need?
Start with roles/mcp.toolUser. It includes mcp.tools.call, the permission needed to invoke MCP tools. Then consult the product reference for each operation’s resource permissions. For example, the IAM MCP server can inspect and manage custom roles and deny policies. Google’s IAM example lists:
roles/mcp.toolUserfor MCP calls;roles/iam.roleAdminfor custom-role management;roles/iam.denyAdminfor deny-policy management.
Those latter permissions can alter access policy. Treat them as privileged production changes, require human approval where appropriate, and avoid granting broad project-owner access merely to make an agent work.
Are Google Cloud MCP tools read-only?
No universal read-only guarantee exists. Tool behavior is product-specific: some tools retrieve information, while others can create, update or delete resources or policies. Read the selected product’s MCP reference, inspect tool descriptions in the client, and map every tool to an IAM permission before enabling it. Use separate identities or projects for experimentation, and place an approval gate in front of destructive or access-changing calls.
Security, governance and Model Armor
Google documents IAM-based fine-grained controls, centralized audit logging and optional Model Armor scanning for MCP calls and responses. Model Armor is not a blanket compliance guarantee. Regional availability and routing can affect where processing occurs, and logging may record the entire payload. Resource/read calls used to render an MCP App are not scanned by Model Armor even when tool calls are scanned. Check the current regional and logging behavior against your data-residency policy before sending sensitive content.
Google-managed remote server versus local or self-hosted MCP
| Decision factor | Google-managed endpoint | Local or self-hosted server |
|---|---|---|
| Operational ownership | Runs on Google’s service infrastructure. | You operate the local process or deployment, upgrades and availability. |
| Coverage | Uses the tools and operations published for that Google product. | Depends on the server implementation and its integrations. |
| Identity | Google authentication plus roles/mcp.toolUser and service IAM. |
Depends on the server’s credential and authorization design. |
| Governance | Review endpoint region, audit behavior and Model Armor availability. | You control hosting location, logging and network controls. |
Troubleshooting common failures
401 or 403 responses
A 401 usually indicates a missing, expired or unsupported credential. A 403 commonly means the identity lacks mcp.tools.call or the underlying product permission. Confirm the principal in the token, grant the minimum required roles at the correct project or resource level, and retry after IAM propagation.
The endpoint works but a tool is missing
Tool availability is service- and preview-dependent. Verify that you selected the correct endpoint, enabled the product API and are using the current product MCP reference. Do not substitute another product’s URL.
Client initialization or session errors
Clients implementing an older MCP handshake may not match the stateless behavior documented for version 2026-07-28. Update the client or follow its Google Cloud integration instructions, and verify whether headers or _meta are required.
Writes are unexpectedly rejected
Read access and MCP invocation do not imply write access. Identify the exact failing operation and grant its documented service permission, preferably on the narrowest resource. If the operation changes IAM or production infrastructure, use an approval workflow instead of widening roles.
Requests contain sensitive data
Review payload logging, endpoint region and Model Armor coverage before transmitting the data. Redact unnecessary fields and ensure audit access is limited to authorized operators.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Or skip the browser setup
If your task is obtaining a clean image or PDF of a web page rather than calling a Google Cloud service, ScreenshotNeo is a separate website screenshot API and MCP server. One GET request returns PNG, JPEG, WebP or PDF. It accepts cookie banners before capture and removes more than 60 known consent platforms, newsletter popups and chat widgets; failed loads, bot checks, CAPTCHAs, blank pages, timeouts and cache hits are not billed, and response headers identify the page verdict and billing status. Its MCP tools—take_screenshot, get_page_info and capture_pdf—work with Claude, Cursor and other MCP clients.
Using the ScreenshotNeo API documentation:
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpPython:
import requests
r = requests.get("https://api.screenshotneo.com/v1/shot", params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"}, timeout=90)
open("shot.webp", "wb").write(r.content)Node.js:
const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);The Free plan includes 1,000 screenshots per month without a card; paid plans start at $5 for 3,000 shots. Create a free ScreenshotNeo account.
Best Value
Operational checklist
- Record the exact product endpoint, region and preview status.
- Enable the product API in the intended project.
- Use a dedicated, narrowly scoped identity.
- Grant
roles/mcp.toolUserand only documented service permissions. - Test read operations before enabling writes.
- Review audit logging, payload exposure and Model Armor regional limits.
- Recheck Google’s catalogue and protocol documentation before production rollout.
Frequently Asked Questions
Do I need one MCP URL for all Google Cloud services?
No. Google publishes product-specific managed endpoints, and each endpoint has its own tools and permissions.
Can a Google Cloud MCP server replace gcloud?
Not automatically. MCP exposes the operations implemented by the selected service server; it is not a universal command-line replacement.
Where should I find the current endpoint list?
Use Google’s live supported-products catalogue and the MCP reference for the specific service, because entries and preview status change.
The Bottom Line
Choose the product-specific endpoint, authenticate with a client-supported Google method, grant roles/mcp.toolUser plus narrowly scoped service permissions, and treat every write-capable tool as a privileged operation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
