Skip to content
English
Go to merchantspring.io
  • There are no suggestions because the search field is empty.

Why can the same MerchantSpring MCP prompt produce different results?

The underlying MerchantSpring data can remain the same while an AI model presents or interprets it differently.

A language model is involved in deciding which tools to use and how to explain the response. Language models can introduce variability, even when the prompt and underlying data are unchanged.

What may change?
  • The wording or length of the summary
  • Which observations the AI application highlights
  • The layout of a table, chart or dashboard
  • How an ambiguous request is interpreted
  • Which tool is selected when several tools appear relevant
What should remain consistent?

When the same endpoint, parameters and processed data are used, the raw MerchantSpring API response should follow the endpoint's defined behaviour. The AI-generated interpretation may still vary.

How can I reduce variation?
  • Use a detailed prompt with explicit filters and output requirements
  • Ask the AI application to state its assumptions
  • Review the underlying tool call or raw response where available
  • Use the direct API for a fixed recurring output
Important to know

MerchantSpring can refine the connector's tools and instructions to reduce common mistakes, but MerchantSpring does not train the external AI model itself.

Need more help? Contact MerchantSpring Support at support@merchantspring.io.