WebMCP Browser Support: Chrome 146, Polyfill, and What Works Today

Native support: Chrome and Edge 146+

Google shipped navigator.modelContextas a stable API in Chrome 146 (released early 2026). Edge follows Chromium's release schedule and shipped the same version.

In these browsers, WebMCP is a first-class browser API — no JavaScript shim, no workaround. AI agents that understand WebMCP can query it directly from the browser's own API surface.

The WebMCP polyfill

For every other browser, the WebMCP polyfill (maintained by mcp-b) implements the same navigator.modelContextAPI in JavaScript. It works by injecting the API surface into the page environment — AI agents that are aware of WebMCP query the same path regardless of whether it's native or polyfilled.

Aigentably loads the polyfill automatically as part of its snippet. You don't configure anything — your tools work in every modern browser on day one.

Browser compatibility table

Firefox and Safari

Mozilla and Apple have not announced implementation timelines for WebMCP. The W3C Web Machine Learning Community Group draft is a joint Google/Microsoft submission — Firefox and Safari have expressed interest but not committed.

This doesn't affect your site. The polyfill covers both. When/if native support ships for those browsers, it will be a transparent upgrade — tool registration code is identical regardless of implementation layer.

What this means for AI agents

AI agent integration with WebMCP depends on the agent, not just the browser. Claude, ChatGPT, and Gemini each have their own integration timelines for querying navigator.modelContext. Chrome shipping the native API accelerates this — it's a signal to agent developers that WebMCP is stable to build on.

For your site: registering tools now means you're ready when agents query for them. There's no cost to being early.

Related