Tool Result

Artifact tools published

Also known as: Function Result, Tool Output, Tool Response

Definition

The output returned after executing a tool invocation, which is then provided back to the LLM for incorporation into its reasoning or response. Tool results close the loop in tool use: model requests tool → application executes → result fed back → model continues. Results may include data, confirmation, or error information.

What this is NOT

  • Not the tool definition (definition is input; result is output)
  • Not the model's final response (result is intermediate; response is what user sees)
  • Not the raw API response (tool result is formatted for the model)

Alternative Interpretations

Different communities use this term differently:

llm-practitioners

The content provided in a tool/function result message that follows a tool call. Typically serialized as a string or JSON, it becomes part of the conversation context for the model's next turn.

Sources: OpenAI Function Calling documentation, Anthropic Tool Use documentation

Examples

  • {"temperature": 72, "conditions": "sunny", "location": "San Francisco"}
  • "Successfully sent email to user@example.com"
  • {"error": "Invalid API key", "code": 401}
  • Search results: 1. [Title](url) - snippet... 2. [Title](url) - snippet...

Counterexamples

Things that might seem like Tool Result but are not:

  • The tool definition (that's input, not output)
  • The model's interpretation of the result (that's reasoning)
  • The final response to the user (that incorporates the result)

Relations

  • requires tool-use (Tool results come from tool use)
  • requires function-calling (Results are returned in function calling flow)
  • overlapsWith agent-loop (Tool results feed back into the agent loop)

See Also