WebMCP proposes structured browser tools to replace brittle DOM scraping by agents
Tara Agyemang from the Google Chrome team introduces WebMCP, a proposed web standard that lets sites expose named, typed actions directly to AI agents instead of forcing them to parse the DOM, accessibility tree, and screenshots.
Score breakdown
WebMCP, if adopted as a web standard, replaces the fragile, token-intensive DOM-scraping approach agents currently use with direct, structured tool calls — reducing the work agents must do to complete actions on existing websites.
- 01WebMCP is a proposed web standard from the Google Chrome team that exposes site capabilities as structured tools for AI agents.
- 02Current agent interactions require parsing the full DOM, accessibility tree, and screenshots — a process described as token-heavy and brittle.
- 03The declarative API adds HTML attributes to existing forms; the browser then auto-generates the JSON schema for agents.
Tara Agyemang, a developer relations engineer on the Google Chrome team, presented WebMCP at AI Engineer as a proposed web standard aimed at reducing the overhead AI agents face when interacting with websites. Today, an agent attempting a simple task like buying two concert tickets must process the entire DOM, parse the accessibility tree, take and analyze a screenshot, calculate pixel coordinates, and attempt a click — a process that is slow, token-expensive, and fragile. A shifted layout caused by a late-loading ad, for instance, can cause the click to miss entirely.
WebMCP addresses this by letting site owners define their site's capabilities as a menu of named, typed, and described tools that agents can call directly.
WebMCP addresses this by letting site owners define their site's capabilities as a menu of named, typed, and described tools that agents can call directly. Agyemang outlined two implementation paths: the declarative API, which adds a small number of HTML attributes to existing forms and lets the browser generate the JSON schema automatically, and the imperative API, which allows developers to register custom tools in JavaScript for complex multi-step flows, including an `execute` block that runs normal DOM code and returns state to the agent. The live demo showed a concert ticket purchase completed in just three tool calls — search by name, open the concert page, and call purchase with quantity and section specified.
Agyemang also emphasized that strong web foundations — semantic HTML, robust accessibility standards, good page performance, and clear user experience flows — are prerequisites that make a site agent-ready by default, and that WebMCP builds on top of those rather than replacing them. WebMCP is still experimental and in early preview on Chrome 146; an eval CLI and inspector extension are currently available for developers to test their own sites.
Key facts
- 01WebMCP is a proposed web standard from the Google Chrome team that exposes site capabilities as structured tools for AI agents.
- 02Current agent interactions require parsing the full DOM, accessibility tree, and screenshots — a process described as token-heavy and brittle.
- 03The declarative API adds HTML attributes to existing forms; the browser then auto-generates the JSON schema for agents.
- 04The imperative API lets developers register custom JavaScript tools with an `execute` block for complex multi-step flows.
- 05A live demo completed a concert ticket purchase in three tool calls: search by name, open the concert page, and call purchase with quantity and section.
- 06WebMCP is in early preview on Chrome 146, with an eval CLI and inspector extension available for testing.
- 07Agyemang noted that improving semantic HTML, accessibility, and page performance are prerequisites before adopting WebMCP.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 12, 2026 · 10:05 UTC. How this works →