Part 2: Tool-Specific Parameters
Tool-specific parameters for the MCP endpoint’s params object
Overview
This section covers the parameters that go inside the params
object of the JSON-RPC request. These parameters vary based on which tool you’re calling through the MCP endpoint.
Common Parameter
name
- Type:
string
- Required: Yes
- Allowed Values:
"chat_gwi"
- For asking questions about audience insights"explore_insight_gwi"
- For getting detailed breakdowns of specific insights
- Description: Specifies which tool to call
Tool 1: Chat GWI
Purpose
The chat_gwi
tool allows you to ask natural language questions about audience behavior, demographics, preferences, and market insights.
Arguments
prompt
- Type:
string
- Required: Yes
- Min Length: 1
- Description: Natural language question about audience behavior, preferences, or market insights
Good questions include:
- Demographics: “What is the age distribution of Tesla owners?”
- Behavior: “How do Gen Z consumers use social media differently?”
- Preferences: “What streaming services do premium car owners prefer?”
- Geographic: “Compare shopping habits between US and UK millennials”
- Trends: “How has remote work adoption changed among tech workers?”
Tips for best results:
- Be specific about the audience segment
- Include geographic location if relevant
- Specify time periods for trend analysis
- Ask one focused question at a time
chat_id
- Type:
string
- Required: No
- Default:
""
- Description: Optional. Use the chat_id from a previous chat_gwi response to continue the conversation and maintain context for follow-up questions.
When to use:
- Asking follow-up questions about the same topic
- Refining or narrowing down previous results
- Comparing different aspects of the same audience
When NOT to use:
- Starting a completely new topic
- Asking about entirely different audience segments
Complete Example
Follow-up Example
Tool 2: Explore Insight GWI
Purpose
The explore_insight_gwi
tool provides detailed breakdowns of a specific insight returned by the chat_gwi
tool.
Arguments
insight_id
- Type:
string
- Required: Yes
- Format: UUID (e.g., “a1b2c3d4-e5f6-7890-abcd-ef1234567890”)
- Pattern:
^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$
- Description: The unique identifier of an insight returned by chat_gwi
What it returns:
- Exact percentages for each answer option
- Sample sizes for statistical reliability
- Index scores showing over/under-indexing vs general population
Important notes:
- Must be a valid UUID from a previous chat_gwi response
- Cannot be used with arbitrary or made-up IDs
- Each insight_id corresponds to a specific data point from the chat response
Complete Example
Usage Scenarios
Scenario 1: New Question
When starting a new conversation about audience insights:
Scenario 2: Follow-up Question
When continuing a conversation to get more related insights:
Scenario 3: Detailed Breakdown
When you need statistical details about a specific insight: