Chrome's Modern Web Guidance 1.0 brings curated web-platform expertise to coding agents

Chrome ships Modern Web Guidance 1.0

Chrome has released version 1.0 of Modern Web Guidance, a toolkit designed to fill gaps in how AI coding agents handle modern web development. The announcement came in a Chrome for Developers blog post dated November 10, 2026, which was surfaced on Hacker News's front page. The project first appeared in early access at Google I/O in May.

Modern Web Guidance is a set of skills — packaged guidance and data installed into a project — so that coding agents draw on vetted web-platform best practices and browser compatibility data when they write code.

Where coding agents fall short

According to Chrome's own evaluations, even leading coding models that reliably produce working code still miss important practices in four areas:

• Accessibility: keyboard focus and inert handling, reduced-motion support, and accessible form validation.
• User experience: scroll and touch gestures, dialog and popover transitions, and light/dark theme implementations.
• Performance: main-thread task scheduling, isolating offscreen rendering, and preventing layout shift.
• Security: safe HTML sanitization, strict Content Security Policies, and passkey credential management.

The Chrome team flags a particular hazard in the first category: bad ARIA usage can leave a page less accessible than if ARIA had never been used at all. User-facing guides in the 1.0 release were reviewed with real assistive technologies by accessibility experts who use those guides daily.

What ships in 1.0

The stable release adds four things:

1. A modernization workflow that scans existing HTML, CSS, JavaScript and package.json files to find upgrade opportunities an agent can apply safely. The first version uses static analysis against a deliberately narrow set of patterns to keep results predictable; Chrome says it plans to broaden coverage and add agentic checks where the agent reasons directly about the code.
2. An expanded guidance database targeting hundreds of verified AI knowledge gaps, with expert-approved cross-browser solutions.
3. Deeper expert review of every guide, including sign-off from technical leads in the relevant content area.
4. Full evaluation coverage across all guides, plus a public evaluation dashboard showing measurable improvements over agents running without the guidance.

Version 1.0 also makes the project a cross-browser effort: Firefox and Edge are joining Chrome to support, improve and contribute to compatibility guidance, building on the shared Baseline initiative.

Platform features over JavaScript

A recurring theme in the announcement is that models default to JavaScript even when the platform already offers a better tool. Native HTML and CSS capabilities for animating popups and dialogs, scroll-driven interactions and smooth page transitions often require less code, perform better because the browser handles them, and carry accessibility for free. Newer capabilities such as passkeys and WebAuthn, and improved autofill for custom form controls on mobile, can change the end-user experience in ways users actually notice.

Because the guidance embeds Baseline compatibility data, recommendations adapt to a project's browser-support requirements rather than suggesting features a given audience cannot yet rely on.

Chrome also notes the guides work as a standalone learning resource, with a CLI for searching use cases, since domain expertise remains valuable even in agent-assisted coding. The team recommends pairing the toolkit with Chrome DevTools for agents, which gives agents runtime visibility into how code behaves across devices, user preferences and network conditions.

Installation is a single command from a project directory: npx modern-web-guidance@latest install, with npx modern-web-guidance@latest update for upgrading earlier installs.

Why it matters

The Chrome team frames the problem bluntly: once every developer has a coding agent, raw output speed stops being a competitive advantage, and quality — usability, accessibility, performance and security — becomes the differentiator. Agents trained on older material systematically underuse modern platform features and reproduce dated patterns. Modern Web Guidance attacks that with curated, maintained, cross-browser-vetted data rather than waiting for base models to catch up, and the arrival of Firefox and Edge as contributors points toward a shared industry layer between models and the web platform. For teams already running agents against real codebases, the new modernization workflow also offers a low-friction way to clear out accumulated polyfills and workarounds.