MCP tools return types

[gilv]

Pre-submission Checklist

• I have verified this would not be more appropriate as a feature request in a specific repository
• I have searched existing discussions to avoid duplicates

Your Idea

I wonder what community thinks if there should be some spec, pattern or perhaps MCP server support related return types from MCP methods. Currently every MCP implementation can return anything that developer decided to return. As example, a simple create Slack channel MCP method might look like


``

def create_channel(ctx: Context, channel_name: str, is_private: bool = False) -> str:
"""Create a new channel by name."""
try:
response = slack_client.conversations_create(name=channel_name, is_private=is_private)
return response["channel"]["id"]
except SlackApiError as e:
return f"Error: `{e.response['error']}"

``

But return can be anything from response[channel] with the whole metadata, response[channel][id] or even any string message with “Channel {channel_name} was created successfully”.

If MCP tool responses are chained into complex flows with multi tool invocations, where tool calling input using outputs from previous methods then seems everyone implements different approaches. I wonder if community thinks there should be spec and perhaps MCP server support to provide a standard approach on tool chaining, where responses of one tool are used as inputs to another. 


Scope

• Protocol Specification
• SDK Features
• Documentation
• Developer Experience
• Other