A Model Context Protocol (MCP) server acts as a proxy between an external service that provides context, data, or capabilities to a Large Language Model (LLM) or AI application. MCP servers connect AI applications to external systems such as databases and web services, translating their responses into a format that the AI application can understand.
Server Setup
You must enable MCP servers and set up authentication before use. For more information about using Google and Google Cloud remote MCP servers, see Google Cloud MCP servers overview .
An MCP server for interacting with Cloud Composer resources.
Server Endpoints
An MCP service endpoint is the network address and communication interface (usually a URL) of the MCP server that an AI application (the Host for the MCP client) uses to establish a secure, standardized connection. It is the point of contact for the LLM to request context, call a tool, or access a resource. Google MCP endpoints can be global or regional.
The composer.googleapis.com MCP server uses only regional endpoints:
- https://composer.{region}.rep.googleapis.com/mcp
MCP Tools
An MCP tool is a function or executable capability that an MCP server exposes to a LLM or AI application to perform an action in the real world.
The composer.googleapis.com MCP server has the following tools:
Initiates the creation of a new Cloud Composer environment with default settings.
This tool creates a small private IP environment with a default configuration, including default networking settings.
Creating public IP environments, or customizing most settings (such as network, subnetwork, network attachment, environment size, CMEK or custom environment bucket) is not supported.
If the user requests settings that aren't supported by the tool parameters, inform them that this tool can't create an environment in the requested configuration, and that they should use gcloud CLI or Cloud Console instead.
Some environment configuration parameters are mandatory. Always ask about them:
-
Always ask the user to provide a service account that they want to use for their environment. Users are strongly recommended to use a dedicated user-managed service account and are discouraged from using the default Cloud Compute service account for security reasons.
-
Always ask the user to specify the IP ranges that they want to use to access the Airflow UI and web server. If the user doesn't specify IP ranges, clarify that the default is not allowing access from any IPs. We also recommend to not allow access from all IP addresses to avoid potential security risks.
This tool creates an environment without pre-installed PyPI packages. Use the manage_pypi_packages
tool to install PyPI packages after the environment is created.
This tool triggers the environment creation process, which is a long-running operation. The tool returns an operation object. Use the get_operation
tool with the operation name returned by this tool to poll for creation status. Environment creation typically takes 25-30 minutes. A completed operation will include the details of the created environment.
Retrieves details about a specific Cloud Composer environment, including its configuration and status.
Use this tool to check the current state of an environment, its Airflow version, Airflow UI URL, environment's bucket, or other configuration details.
Lists all Cloud Composer environments in a given project and location.
Use this tool to when you need to know which environments exist in a specific project and location.
Results are paginated; use page_size
to control the number of results per page and page_token
from a previous response to get the next page. If page_token
is not present in the response, there are no more results.
Initiates the deletion of a Cloud Composer environment.
Warning: This is a destructive action permanently deletes an environment and can't be undone. Always ask the user to confirm the deletion before proceeding.
This tool triggers the environment deletion process, which is a long-running operation that typically takes 10-20 minutes. The tool returns an operation object. Use the get_operation
tool with the operation name returned by this tool to poll for deletion status.
Manages PyPI packages in a Cloud Composer environment.
Use this tool to install and uninstall PyPI packages in the Cloud Composer environment, or change PyPI package versions or extras.
To install packages or change versions or extras, use the update_packages
field. To remove packages, use the remove_packages
field. At least one package must be specified in one of these fields. The same package can't be specified in both fields.
This tool triggers an environment update, which is a long-running operation that can take several minutes. The tool returns an operation object. Use the get_operation
tool to poll its status. A completed operation will include the details of the updated environment.
Gets the latest status of a long-running operation.
Several tools like create_environment
, delete_environments
, and manage_pypi_packages
return an operation object, because they initiate processes that take time to complete. Use this tool to check if such an operation has finished and whether it was successful.
Provide the name
of the operation as returned by the tool that initiated it.
A completed operation may include the details specific to the operation, described in the description of the tool that initiates the operation.
Retrieves the DAG source code from a Cloud Composer environment.
Use this tool when you want to understand DAG implementation details, for example to understand why a DAG run has failed, or to debug DAG's behavior. The tool returns the latest version of the source code.
Retrieves details of a specific DAG run from a Cloud Composer environment.
Use this tool when you need to retrieve details of a specific DAG run, for example its state, start time or end time.
Retrieves details of a specific task instance from a Cloud Composer environment.
Use this tool to retrieve details of a specific task instance, such as its state, start time, end time, or parameters, to understand its status or to debug a failure.
This tool does not return task instance logs directly. Instead, use the cloud_logging_filter
field from the response to query Cloud Logging for the logs for this task instance.
Finds failed DAG runs in a Cloud Composer environment.
Use this tool to find recent DAG run failures, for example to begin troubleshooting.
If the user wants to troubleshoot a specific DAG, provide the DAG ID to get the details of its last failed run. If the user doesn't know which DAG is failing, or wants to see recent failures across all DAGs, leave the DAG ID empty; in this case, the tool will return the last failed DAG run for every DAG in the environment that has at least one failure. If the response returns more than one DAG run and the user asked for troubleshooting or explaining the failure, ask the user which DAG run to troubleshoot.
Lists failed task instances of a specific DAG run in a Cloud Composer environment.
If a DAG run has failed, use this tool to get the list of task instances of this DAG run that are in the FAILED state and may have caused the DAG run failure.
This tool does not return task instance logs directly. Instead, for each task instance, use the cloud_logging_filter
field from the response to query Cloud Logging for the logs for that task instance.
Get MCP tool specifications
To get the MCP tool specifications for all tools in an MCP server, use the tools/list
method. The following example demonstrates how to use curl
to list all tools and their specifications currently available within the MCP server.
| Curl Request |
|---|
curl --location 'https://composer.{region}.rep.googleapis.com/mcp' \ --header 'content-type: application/json' \ --header 'accept: application/json, text/event-stream' \ --data '{ "method": "tools/list", "jsonrpc": "2.0", "id": 1 }' |
