Add timeout and error handling for async external service calls in toolsets

[jmhbh]

Is your feature request related to a problem? Please describe.
In mcp_toolset.py if the get_tools() method does not receive a response due to an issue with the MCP server then this can hang forever until a top level timeout kicks in. However, since there is no timeout or any error handling, the user is in the dark about the error that occurred. There should be better error handling and possibly a configurable/default timeout when handling async calls to external services.

Describe the solution you'd like
Expose a configurable timeout or at least implement a default timeout for async calls to external services alongside appropriate error messages that will inform the user where exactly a failure occurred.

[Jacksunwei]

I think this is more timeout config that should be set on MCP Session.

@jmhbh have you tried whether MCP client/server can take a timeout param?

[kevinkassimo]

@Jacksunwei I'll take a look at this

[kevinkassimo]

I think timeout config is already exist for StdioConnectionParams, SseConnectionParams and StreamableHTTPConnectionParam (see timeout param):

adk-python/src/google/adk/tools/mcp_tool/mcp_session_manager.py

Lines 53 to 108 in 87dcb3f

	class StdioConnectionParams(BaseModel):
	"""Parameters for the MCP Stdio connection.
	Attributes:
	server_params: Parameters for the MCP Stdio server.
	timeout: Timeout in seconds for establishing the connection to the MCP
	stdio server.
	"""
	server_params: StdioServerParameters
	timeout: float = 5.0
	class SseConnectionParams(BaseModel):
	"""Parameters for the MCP SSE connection.
	See MCP SSE Client documentation for more details.
	https://github.com/modelcontextprotocol/python-sdk/blob/main/src/mcp/client/sse.py
	Attributes:
	url: URL for the MCP SSE server.
	headers: Headers for the MCP SSE connection.
	timeout: Timeout in seconds for establishing the connection to the MCP SSE
	server.
	sse_read_timeout: Timeout in seconds for reading data from the MCP SSE
	server.
	"""
	url: str
	headers: dict[str, Any] | None = None
	timeout: float = 5.0
	sse_read_timeout: float = 60 * 5.0
	class StreamableHTTPConnectionParams(BaseModel):
	"""Parameters for the MCP Streamable HTTP connection.
	See MCP Streamable HTTP Client documentation for more details.
	https://github.com/modelcontextprotocol/python-sdk/blob/main/src/mcp/client/streamable_http.py
	Attributes:
	url: URL for the MCP Streamable HTTP server.
	headers: Headers for the MCP Streamable HTTP connection.
	timeout: Timeout in seconds for establishing the connection to the MCP
	Streamable HTTP server.
	sse_read_timeout: Timeout in seconds for reading data from the MCP
	Streamable HTTP server.
	terminate_on_close: Whether to terminate the MCP Streamable HTTP server
	when the connection is closed.
	"""
	url: str
	headers: dict[str, Any] | None = None
	timeout: float = 5.0
	sse_read_timeout: float = 60 * 5.0
	terminate_on_close: bool = True

The only one without it is from mcp package's own StdioServerParameters, but StdioServerParameters is already a wrapper of it to add timeout field.

Have you tried setting them with a small value and see if the timeout is enforced for your case?