microsoft/semantic-kernel · error · ServiceInvalidExecutionSettingsError

Tool choice 'none' is not supported by Anthropic.

Error message

Tool choice 'none' is not supported by Anthropic.

What it means

AnthropicChatPromptExecutionSettings.validate_tool_choice (a pydantic model_validator) raises ServiceInvalidExecutionSettingsError when tool_choice is present and its type equals FunctionChoiceType.NONE. The Anthropic Messages API has no 'none' tool-choice mode, so the connector rejects the configuration before any request.

Source

Thrown at python/semantic_kernel/connectors/ai/anthropic/prompt_execution_settings/anthropic_prompt_execution_settings.py:53

            description=(
                "Do not set this manually. It is set by the service based on the function choice configuration."
            ),
        ),
    ] = None
    tool_choice: Annotated[
        dict[str, str] | None,
        Field(
            description="Do not set this manually. It is set by the service based on the function choice configuration."
        ),
    ] = None

    @model_validator(mode="after")
    def validate_tool_choice(self) -> "AnthropicChatPromptExecutionSettings":
        """Validate tool choice. Anthropic doesn't support NONE tool choice."""
        tool_choice = self.tool_choice

        if tool_choice and tool_choice.get("type") == FunctionChoiceType.NONE.value:
            raise ServiceInvalidExecutionSettingsError("Tool choice 'none' is not supported by Anthropic.")

        return self

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Use AUTO or REQUIRED for Anthropic, or omit tool_choice entirely.
2. To avoid tool use, simply do not register/pass any tools rather than selecting NONE.
3. Branch the function-choice behavior per provider so NONE is never sent to Anthropic.

Example fix

// before
settings = AnthropicChatPromptExecutionSettings()
settings.tool_choice = {"type": "none"}  # rejected

// after
settings.tool_choice = {"type": "auto"}
# or leave tool_choice unset and provide no tools to suppress tool use

from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceType
tc = settings.tool_choice
if tc and tc.get('type') == FunctionChoiceType.NONE.value:
    raise ValueError('Anthropic does not support tool_choice NONE; use AUTO/REQUIRED or omit tools')

def is_anthropic_compatible_tool_choice(tool_choice: dict | None) -> bool:
    if not tool_choice:
        return True
    return tool_choice.get('type') != FunctionChoiceType.NONE.value

When it happens

Trigger: Configuring an Anthropic chat with function_choice_behavior/tool_choice set to NONE, or passing tool_choice={"type": "none"} in the execution settings.

Common situations: Sharing/reusing a function-choice configuration across providers and carrying a NONE setting onto Anthropic; explicitly disabling tools by choosing NONE; defaulting to NONE in a generic helper.

Related errors

• Failed to create Anthropic settings.
• The Anthropic chat model ID is required.
• Tool message found without a preceding message.
• Tool message found after a user or system message.
• Ollama does not support function choice behavior of type 're

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/08548eb6e11b52a7. Report an issue: GitHub.