Technical Query: Client-Side Execution vs. Remote API Access in WebMCP

Saurav Gupta

unread,

Feb 27, 2026, 10:24:54 AMFeb 27

to Chrome Built-in AI Early Preview Program Discussions

Question: Since WebMCP is a proposed standard for in-browser agentic capabilities , is there any mechanism to invoke tools registered via the navigator.modelContext API headlessly from an external system, or does the protocol strictly mandate an active browsing context and visible UI for all tool executions?

Best Codes (BestCodes)

unread,

Feb 27, 2026, 10:35:42 AMFeb 27

to Chrome Built-in AI Early Preview Program Discussions, Saurav Gupta

I have the same basic question (https://groups.google.com/a/chromium.org/g/chrome-ai-dev-preview-discuss/c/AHZcNz018ZU). Would love to hear from the team on this!

François Beaufort

unread,

Mar 2, 2026, 4:26:59 AMMar 2

to Chrome Built-in AI Early Preview Program Discussions, bestcodes...@gmail.com, sauravg...@gmail.com

WebMCP folks are discussing whether exposing executeTool() to WebMCP would make sense at https://github.com/webmachinelearning/webmcp/issues/51#issuecomment-3975725141

Reply all

Reply to author

Forward