Developer builds MCP bridge to test Google's experimental WebMCP browser spec

·2 views

A Japan-based developer named tanahiro2010 has built an open-source bridge tool to experiment with WebMCP, a draft specification published by the W3C Web Machine Learning Community Group. WebMCP allows web pages to declare their own features as callable tools for AI agents directly within the browser, using a JavaScript API closely resembling the existing Model Context Protocol (MCP). Unlike MCP, where tools remain registered for the duration of an agent's session, WebMCP tools are only available while the relevant browser tab is open. As of August 2026, the spec remains in an early draft stage and is available for testing via Chrome's Origin Trial from version 149 onward, with the API still subject to change. The developer noted a practical gap between the spec's browser-bound design and real-world use cases such as in-person hands-on sessions where native browser support cannot be guaranteed.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Developer Recreates Georgian Cheese Bread Khachapuri Using Only HTML and CSS

A developer submitted a pure CSS art piece depicting Imeretian Khachapuri, a traditional cheese bread from Georgia's Imereti region, as part of DEV Community's Frontend Challenge Comfort Food Edition. The artwork uses no images, SVG, canvas, or JavaScript — every visual element, including bread texture, steam, and a slicing animation, is built entirely with HTML and CSS. The bread is split into 13 segments using clip-path polygons, with layered radial gradients simulating baked spots, golden toasting, and a buttery gloss. The project also incorporates accessibility features, including reduced-motion support via a media query and correct language tagging for the Georgian closing phrase so screen readers can pronounce it accurately.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Portal data reveals why players quit before a game even loads

A self-published analysis by game portal SlowDen, conducted on 3 August 2026, examined why visitors leave game pages without ever pressing play. Measurements taken on three live games showed that on a typical phone viewport, the game box was just 330 by 205 pixels — roughly a quarter of the screen — making title screens built for larger canvases nearly illegible. On common laptop viewports around 1536 by 695 pixels, up to a third of the game box fell below the visible fold on arrival. The study also recorded several seconds of black screen before any content rendered, and a start button positioned at the very edge of the visible area. The portal notes that this kind of host-side data is rarely available to developers, and published its own figures precisely because the shortcomings identified belong to the host itself.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Student Essay Argues Heroism Is Within Everyone's Reach

A student essay by Brian Kim, written for a World Masterpieces class, explores the evolving concept of heroism beyond traditional superhero tropes. Drawing on 'The Epic of Gilgamesh,' Kim examines how characters like Gilgamesh and Enkidu demonstrate that heroes emerge through their actions and growth, not just innate abilities. The essay references the concept of the 'Banality of Heroism,' which holds that anyone can perform a heroic act when the moment calls for it. Kim also uses the film 'Hancock' to illustrate how societal acceptance and self-identity shape whether someone is perceived as a hero. Ultimately, the essay concludes that overcoming personal obstacles and choosing to act for others is what defines a true hero.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Snowflake Launches Per-User Quota Feature to Control Individual Credit Spending

Snowflake has moved its Per-User Quota feature into Public Preview, allowing account administrators to set credit limits for individual users across compute and AI services. The feature addresses growing concerns around a small number of users consuming disproportionate shares of warehouse or AI credits, often without timely visibility. Administrators can apply limits monthly for warehouse usage, and monthly, daily, or both for AI-related services including AI Functions, Cortex Agents, CoCo, and Snowflake CoWork. Users can be grouped into named quotas and targeted via tags, enabling different spending tiers such as standard and power-user allocations. When limits are breached, AI requests can be automatically blocked, while warehouse overages can trigger a custom stored procedure defined by the administrator.

0 comments Read more at DEV Community