[Model Context Protocol (MCP)](https://modelcontextprotocol.io/) servers let your AI applications connect
with databases, calendars, productivity tools, and APIs. You can list available
Google and Google Cloud remote MCP servers, authenticate your AI applications,
and configure optional security and safety settings.

You can use MCP servers when building AI applications or when working with AI
assistants such as Claude Code, Gemini Code Assist, Gemini
CLI, and Google AI Studio. Because Google and Google Cloud MCP servers run on
Google infrastructure instead of your local machine, their endpoints are
available as soon as you enable the service API. For more information, see
[Google Cloud MCP servers overview](https://docs.cloud.google.com/mcp/overview.md.txt).

## Before you begin

<br />

### Required roles

[Non-Google Cloud MCP servers](https://docs.cloud.google.com/mcp/supported-products.md.txt#google-mcp-servers) don't require the MCP Tool
User (`roles/mcp.toolUser`) role. You also can't use Identity and Access Management (IAM)
policies or permissions to control access to non-Google Cloud MCP servers.


To get the permissions that
you need to use Google Cloud MCP servers,

ask your administrator to grant you the
[MCP Tool User](https://docs.cloud.google.com/iam/docs/roles-permissions/mcp.md.txt#mcp.toolUser) (`roles/mcp.toolUser`) IAM role on the Google Cloud project.


For more information about granting roles, see [Manage access to projects, folders, and organizations](https://docs.cloud.google.com/iam/docs/granting-changing-revoking-access.md.txt).


This predefined role contains

the permissions required to use Google Cloud MCP servers. To see the exact permissions that are
required, expand the **Required permissions** section:


#### Required permissions

The following permissions are required to use Google Cloud MCP servers:

- `mcp.tools.call`
- `resourcemanager.projects.get`
- `resourcemanager.projects.list`


You might also be able to get
these permissions
with [custom roles](https://docs.cloud.google.com/iam/docs/creating-custom-roles.md.txt) or
other [predefined roles](https://docs.cloud.google.com/iam/docs/roles-overview.md.txt#predefined).

## Available MCP servers

Available Google and Google Cloud remote MCP servers are listed in
[Supported products](https://docs.cloud.google.com/mcp/supported-products.md.txt) and are registered in
[Agent Registry](https://docs.cloud.google.com/agent-registry/overview.md.txt) when you enable the product
or service.

## Manage MCP with Agent Registry

To list your available MCP servers and get updated tool descriptions, see
[Manage MCP tools in Agent Registry](https://docs.cloud.google.com/agent-registry/manage-mcp-tools.md.txt).

## List available tools, prompts, and resources

Most MCP clients automatically list available tools, prompts, and resources. The
following tabs describe how to manually list them using HTTP requests:

### Tools

To get a list of available tools and their descriptions, MCP clients use the
MCP [tools/list](https://modelcontextprotocol.io/specification/2026-07-28/server/tools#listing-tools) method in an HTTP request to the MCP server
or toolset endpoint. Authentication isn't required for the `tools/list`
method.

    POST /TOOLSET_ENDPOINT HTTP/1.1
    Host: SERVICE_NAME
    Content-Type: application/json

    {
      "jsonrpc": "2.0",
      "id": 1,
      "method": "tools/list",
      "params": {
        "cursor": "OPTIONAL_CURSOR_VALUE",
        "_meta": {
          "io.modelcontextprotocol/protocolVersion": "MCP_PROTOCOL_VERSION",
          "io.modelcontextprotocol/clientCapabilities": {
            "extensions": {
              "io.modelcontextprotocol/ui": {
                "mimeTypes": ["text/html;profile=mcp-app"]
              }
            ...
            }
          }
        }
      }
    }

Replace the following:

- <var translate="no">`SERVICE_NAME`</var>: The service name---for example, the BigQuery service name is `bigquery.googleapis.com`. For information about listing enabled services, see [List services](https://docs.cloud.google.com/service-usage/docs/list-services.md.txt). To see a list of Google and Google Cloud services that have remote MCP servers without running commands, see [Supported products](https://docs.cloud.google.com/mcp/supported-products.md.txt).
- <var translate="no">`TOOLSET_ENDPOINT`</var>: The remainder of the MCP endpoint after the service name. For example, for Gemini Enterprise Agent Platform, this might be `mcp/models` or `mcp/generate`.
- `OPTIONAL_CURSOR_VALUE`: The cursor value from the previous response, used to retrieve the next page of results.
- `MCP_PROTOCOL_VERSION`: The MCP protocol version. For example, `2026-07-28`.

You can send an HTTP request to list tools directly to a Google remote MCP
server with the following command:

```bash
curl -X POST https://SERVICE_NAME/TOOLSET_ENDPOINT \
    -H 'Content-Type: application/json' \
    -H 'Accept: application/json' \
    -H 'MCP-Protocol-Version: 2026-07-28' \
    -H 'Mcp-Method: tools/list' \
    -d '{
      "jsonrpc": "2.0",
      "id": 1,
      "method": "tools/list",
      "params": {
        "_meta": {
          "io.modelcontextprotocol/protocolVersion": "2026-07-28",
          "io.modelcontextprotocol/clientCapabilities": {
            "extensions": {
              "io.modelcontextprotocol/ui": {
                "mimeTypes": ["text/html;profile=mcp-app"]
              }
            }
          }
        }
      }
    }'
```

The response resembles the following:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "tools": [
      {
        "name": "list_topics",
        "title": "List Topics",
        "description": "Get list of available topics.",
        "input_schema": {...}
        "output_schema": {...}
      },
      {
        "name": "get_topic",
        "title": "Get Topic",
        "description": "Get messages in a Topic.",
        "input_schema": {...}
        "output_schema": {...}
      },
      "nextCursor": "next-page-cursor",
      "ttlMs": 300000,
      "cacheScope": "public"
      ...
    ]
  }
}
```

For more information about the structure of the response, see the MCP
specification for [tools](https://modelcontextprotocol.io/specification/2026-07-28/server/tools), [caching](https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/caching), and
[pagination](https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/pagination).

### Prompts

To get a list of available prompts and their descriptions, MCP clients use
the MCP [prompts/list](https://modelcontextprotocol.io/specification/2026-07-28/server/prompts#listing-prompts) method in an HTTP request to the MCP
server or toolset endpoint. Authentication isn't required for the
`prompts/list` method. All prompts are listed for all toolsets, even if they
don't apply to the toolset that you're using.

    POST /TOOLSET_ENDPOINT HTTP/1.1
    Host: SERVICE_NAME
    Content-Type: application/json

    {
      "jsonrpc": "2.0",
      "id": 1,
      "method": "prompts/list",
      "params": {
        "cursor": "OPTIONAL_CURSOR_VALUE",
        "_meta": {
          "io.modelcontextprotocol/protocolVersion": "MCP_PROTOCOL_VERSION",
          "io.modelcontextprotocol/clientCapabilities": {
            "extensions": {
              "io.modelcontextprotocol/ui": {
                "mimeTypes": ["text/html;profile=mcp-app"]
              }
            }
          }
        }
      }
    }

Replace the following:

- <var translate="no">`SERVICE_NAME`</var>: The service name---for example, the BigQuery service name is `bigquery.googleapis.com`. For information about listing enabled services, see [List services](https://docs.cloud.google.com/service-usage/docs/list-services.md.txt). To see a list of Google and Google Cloud services that have remote MCP servers without running commands, see [Supported products](https://docs.cloud.google.com/mcp/supported-products.md.txt).
- <var translate="no">`TOOLSET_ENDPOINT`</var>: The remainder of the MCP endpoint after the service name. For example, for Gemini Enterprise Agent Platform, this might be `mcp/models` or `mcp/generate`.
- `OPTIONAL_CURSOR_VALUE`: The cursor value from the previous response, used to retrieve the next page of results.
- `MCP_PROTOCOL_VERSION`: The MCP protocol version. For example, `2026-07-28`.

You can send an HTTP request to list prompts directly to a Google remote MCP
server with the following command:

```bash
curl -X POST https://SERVICE_NAME/TOOLSET_ENDPOINT \
    -H 'Content-Type: application/json' \
    -H 'Accept: application/json' \
    -H 'MCP-Protocol-Version: 2026-07-28' \
    -H 'Mcp-Method: prompts/list' \
    -d '{
      "jsonrpc": "2.0",
      "id": 1,
      "method": "prompts/list",
      "params": {
        "_meta": {
          "io.modelcontextprotocol/protocolVersion": "2026-07-28",
          "io.modelcontextprotocol/clientCapabilities": {
            "extensions": {
              "io.modelcontextprotocol/ui": {
                "mimeTypes": ["text/html;profile=mcp-app"]
              }
            }
          }
        }
      }
    }'
```

> [!NOTE]
> **Note:** Not all MCP servers support prompts. Calling `prompts/list` on a server that does not support prompts results in an error.

The response resembles the following:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "prompts": [
      {
        "name": "deploy",
        "description": "Deploys the current working directory to Cloud Run.",
        "arguments": [
          {
            "name": "name",
            "description": "Name of the Cloud Run service to deploy to.",
            "required": false
          }
        ]
      },
      "nextCursor": "next-page-cursor",
      "ttlMs": 600000,
      "cacheScope": "public"
      ...
    ]
  }
}
```

For more information about the structure of the response, see the MCP
specification for [prompts](https://modelcontextprotocol.io/specification/2026-07-28/server/prompts), [caching](https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/caching), and
[pagination](https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/pagination).

### Resources

To get a list of available resources and their descriptions, MCP clients use
the MCP [resources/list](https://modelcontextprotocol.io/specification/2026-07-28/server/resources#listing-resources) method in an HTTP request to the
MCP server or toolset endpoint. Authentication isn't required for the
`resources/list` method. All resources are listed for all toolsets.

    POST /TOOLSET_ENDPOINT HTTP/1.1
    Host: SERVICE_NAME
    Content-Type: application/json

    {
      "jsonrpc": "2.0",
      "id": 1,
      "method": "resources/list",
      "params": {
        "cursor": "OPTIONAL_CURSOR_VALUE",
        "_meta": {
          "io.modelcontextprotocol/protocolVersion": "MCP_PROTOCOL_VERSION",
          "io.modelcontextprotocol/clientCapabilities": {
            "extensions": {
              "io.modelcontextprotocol/ui": {
                "mimeTypes": ["text/html;profile=mcp-app"]
              }
            }
          }
        }
      }
    }

Replace the following:

- <var translate="no">`SERVICE_NAME`</var>: The service name---for example, the BigQuery service name is `bigquery.googleapis.com`. For information about listing enabled services, see [List services](https://docs.cloud.google.com/service-usage/docs/list-services.md.txt). To see a list of Google and Google Cloud services that have remote MCP servers without running commands, see [Supported products](https://docs.cloud.google.com/mcp/supported-products.md.txt).
- <var translate="no">`TOOLSET_ENDPOINT`</var>: The remainder of the MCP endpoint after the service name. For example, for Gemini Enterprise Agent Platform, this might be `mcp/models` or `mcp/generate`.
- `OPTIONAL_CURSOR_VALUE`: The cursor value from the previous response, used to retrieve the next page of results.
- `MCP_PROTOCOL_VERSION`: The MCP protocol version. For example, `2026-07-28`.

You can send an HTTP request to list resources directly to a Google remote
MCP server with the following command:

> [!NOTE]
> **Note:** Not all MCP servers support resources. Calling `resources/list` on a server that does not support resources results in an error.

```bash
curl -X POST https://SERVICE_NAME/TOOLSET_ENDPOINT \
    -H 'Content-Type: application/json' \
    -H 'Accept: application/json' \
    -H 'MCP-Protocol-Version: 2026-07-28' \
    -H 'Mcp-Method: resources/list' \
    -d '{
      "jsonrpc": "2.0",
      "id": 1,
      "method": "resources/list",
      "params": {
        "_meta": {
          "io.modelcontextprotocol/protocolVersion": "2026-07-28",
          "io.modelcontextprotocol/clientCapabilities": {
            "extensions": {
              "io.modelcontextprotocol/ui": {
                "mimeTypes": ["text/html;profile=mcp-app"]
              }
            }
          }
        }
      }
    }'
```

The response resembles the following:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "resources": [
      {
        "uri": "file:///project/src/main.rs",
        "name": "main.rs",
        "title": "Rust Software Application Main File",
        "description": "Primary application entry point",
        "mime_type": "text/x-rust"
      },
      "nextCursor": "next-page-cursor",
      "ttlMs": 300000,
      "cacheScope": "public"
      ...
    ]
  }
}
```

For more information about the structure of the response, see the MCP
specification for [resources](https://modelcontextprotocol.io/specification/2026-07-28/server/resources), [caching](https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/caching), and
[pagination](https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/pagination).

## Authenticate your AI application

Once you have enabled a product or service that supports MCP in your project,
you can use your Google credentials to authenticate and use its tools through
your AI application. For safety and observability, you might want to create a
separate identity with specific permissions for your application instead of
using your own permissions if you are working with production resources. For
more information, see [Authenticate to MCP servers](https://docs.cloud.google.com/mcp/authenticate-mcp.md.txt).

## Optional security and safety configurations

MCP tools can take a wide variety of actions on behalf of AI applications,
introducing security risks and considerations. To help you minimize and manage
these risks, Google Cloud provides default and customizable policies
that control how MCP tools are used within your Google Cloud organization
or project. For more information about MCP security and governance, see
[AI security and safety](https://docs.cloud.google.com/mcp/ai-security-safety.md.txt).

### Enable Model Armor

[Model Armor](https://docs.cloud.google.com/model-armor/overview.md.txt) is a
Google Cloud service designed to enhance the security and
safety of your AI applications. It works by proactively screening LLM prompts
and responses, protecting against various risks and supporting responsible AI
practices. Whether you are deploying AI in your cloud environment, or on
external cloud providers, Model Armor can help
you prevent malicious input, verify content safety, protect sensitive data,
maintain compliance, and enforce your AI safety and security policies
consistently across your diverse AI landscape.

When Model Armor is enabled with
[logging enabled](https://docs.cloud.google.com/model-armor/configure-logging.md.txt), Model Armor logs the entire
payload. This might expose sensitive information in your logs.

#### MCP request routing to Model Armor

Model Armor is available in [certain regions](https://docs.cloud.google.com/model-armor/locations.md.txt). When Model Armor is enabled and you use an MCP server in a jurisdiction that Model Armor doesn't support, the routing behavior of the call might be different for different MCP servers and might break data residency compliance for in-use and in-transit data. For more information about the behavior of individual MCP servers, see [Model Armor supported products](https://docs.cloud.google.com/mcp/model-armor-supported-products.md.txt).

To enable Model Armor, complete the steps in
[Configure Model Armor protection for Google and Google Cloud MCP servers](https://docs.cloud.google.com/model-armor/model-armor-mcp-google-cloud-integration.md.txt).

#### Model Armor logging

For information about Model Armor audit and platform logs, see
[Model Armor audit logging](https://docs.cloud.google.com/model-armor/audit-logging-model-armor.md.txt).

## Use MCP servers

To use Google Cloud remote MCP servers, your administrator must grant the
principal for your AI application the MCP Tool User (`roles/mcp.toolUser`) role
and the required IAM roles for the Google Cloud service
that the tool accesses. [Non-Google Cloud MCP servers](https://docs.cloud.google.com/mcp/supported-products.md.txt#google-mcp-servers)
don't require the MCP Tool User role. You also can't use IAM
policies or permissions to control access to non-Google Cloud MCP servers.

To grant the MCP Tool User role, run the following command:

```bash
gcloud projects add-iam-policy-binding PROJECT_ID \
    --member="PRINCIPAL" \
    --role="roles/mcp.toolUser"
```

Replace the following:

- <var translate="no">`PROJECT_ID`</var>: the Google Cloud project.
- <var translate="no">`PRINCIPAL`</var>: the [principal identifier](https://docs.cloud.google.com/iam/docs/principal-identifiers.md.txt) for the identity you are granting the MCP Tool User role to. For information about the types of principals available on Google Cloud, see [IAM principals](https://docs.cloud.google.com/iam/docs/principals-overview.md.txt).

## What's next

- Learn about [access control with IAM](https://docs.cloud.google.com/mcp/access-control.md.txt).