Google Chrome rolls out WebMCP preview, turning every site into an AI agent tool

Google Chrome’s new WebMCP preview marks a decisive shift from the brittle “scrape‑and‑click” tactics that have long plagued AI agents on the open web. In a blog post on the Chrome for Developers site, Google explains that WebMCP (Web Model Context Protocol) gives sites a standardized way to expose “structured tools” that agents can invoke directly, bypassing the need to infer actions from raw DOM trees [Chrome for Developers, Feb 10 2026]. The preview introduces two complementary APIs: a Declarative API that lets developers declare simple, form‑based actions—such as booking a flight or submitting a support ticket—in HTML, and an Imperative API for more dynamic interactions that require JavaScript execution [Chrome for Developers]. By embedding these capabilities in the browser, Chrome becomes the bridge that translates an agent’s intent into concrete site actions with “increased speed, reliability, and precision,” according to the same announcement.

The practical impact is immediate for developers building AI‑driven products. As Midas Tools reported, WebMCP offers a “third path” between fragile screen‑scraping and heavyweight custom APIs [Midas Tools, Mar 2 2026]. Instead of maintaining a bespoke API for each partner, a site can opt‑in to being “agent‑ready” by defining its capabilities in the new format, and any Chrome‑based agent will automatically understand how to interact with it. VentureBeat’s Sam Witteveen highlighted that this could turn “every website into a structured tool for AI agents,” eliminating the need for per‑site integration work and reducing the maintenance burden when front‑end code changes [VentureBeat]. For enterprises, the promise is that agents can now execute multi‑step workflows—like configuring a checkout flow or filing a detailed support ticket—without the latency and error‑prone heuristics that currently plague automation scripts.

Early adopters are already sketching concrete use cases. In the Chrome blog, Google cites customer‑support scenarios where agents can auto‑populate ticket fields with technical details, e‑commerce flows where agents can locate products, adjust options, and complete checkout with “precision,” and travel bookings that let agents select exact flight itineraries [Chrome for Developers]. These examples illustrate how the Declarative API can handle straightforward form submissions, while the Imperative API unlocks more complex, JavaScript‑driven interactions that were previously inaccessible to agents. The dual‑API design mirrors the “agentic web” vision that Google has been promoting, positioning the browser as the universal interpreter of site‑provided intent rather than a passive renderer.

Analysts see WebMCP as a strategic move to cement Chrome’s relevance in the rapidly expanding AI‑agent ecosystem. Carl Franzen at VentureBeat noted that Google’s “Interactions API”—the umbrella term for WebMCP’s capabilities—could become the de‑facto standard for agent‑site communication, much as the Fetch API standardized network requests a decade ago [VentureBeat]. By baking the protocol into the browser, Google sidesteps the fragmentation that has plagued third‑party extensions and proprietary agent frameworks, offering a single, open surface that developers can rely on across the entire Chrome user base. This could also pressure competing browsers to adopt compatible implementations, lest they lose developer mindshare in the burgeoning “agent‑first” market.

The preview is still early, and Google cautions that WebMCP is “available for early preview” and may evolve before a stable release [Chrome for Developers]. Nonetheless, the announcement signals a clear intent: to transform the web from a passive collection of pages into an active participant in AI‑driven workflows. If the adoption curve follows the pattern of previous web standards, we could see a rapid migration of high‑traffic sites to expose their capabilities via WebMCP, giving AI agents a reliable, native interface to the entire internet. For developers, the takeaway is simple—start experimenting with the Declarative and Imperative APIs now, or risk building on a foundation that may soon be obsolete.