Skip to content

Clarify backend-inference, browser-executed tools as a WebMCP use case #343

Open

Description

@erdalkaraca

Our app provides a chat UI in the browser, while inference runs on our backend. The frontend sends the conversation and available WebMCP tool descriptions to the backend. When the backend returns a tool call, the frontend resolves it to a currently registered browser tool, executes it, and sends the result back for the next inference step.

Is this flow an intended use case for WebMCP’s in-page agent APIs? The existing APIs appear to provide browser-side discovery and execution, while leaving chat and backend orchestration to the application. Would it be useful to clarify this pattern in the explainer or add a non-normative example? We are not proposing a new relay protocol at this stage.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions