Skip to content

Support returning multimodal content such as images from MCP tool calls. #2148

@sandsc

Description

@sandsc

I noticed that FunctionTool already supports returning multimodal results like ToolOutputImage, but currently invoke_mcp_tool in file util.py converts all content into strings.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions