Philip Walton announced Modern Web Guidance 1.0 on the Chrome for Developers blog on November 10, 2026. Chrome showed an early preview at Google I/O in May. Version 1.0 is the stable release, and Firefox and Edge now contribute to the project alongside Chrome.
Modern Web Guidance is a set of skills you install into a coding agent. Each guide carries web platform best practices plus Baseline browser-compatibility data, so the agent can pick a feature that fits your support targets.
# new install, from the project directory
npx modern-web-guidance@latest install
# already on the preview
npx modern-web-guidance@latest update
You can also install it from the plugin marketplaces in Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build.
What 1.0 adds
The post lists four changes over the preview:
| Area | 1.0 |
|---|---|
| Modernization | New workflow scans an existing codebase for safe upgrades |
| Guides | Hundreds of verified AI knowledge gaps covered |
| Review | Tech-lead sign-off on each guide; accessibility review on user-facing guides |
| Evals | Full coverage with a public dashboard comparing agents with and without the guides |
The Chrome team names the areas where frontier models still miss. In accessibility: keyboard focus and inert management, reduced-motion support, accessible form validation. In performance: main-thread task scheduling and layout shift prevention. In security: HTML sanitization, strict Content Security Policies and passkey credential management. Accessibility specialists who use assistive technology every day tested the user-facing guides for 1.0.
The modernize workflow
The new workflow reads your HTML, CSS, JavaScript and package.json, looks for outdated workarounds, dependencies you no longer need and dead polyfills, and walks the agent through the changes. You call it from the agent:
/modern-web-guidance Modernize the code in this directory
Keep your expectations narrow for now. The first release uses static analysis and checks a limited set of patterns. Chrome says it picked a small scope for predictable results and plans to add agentic audits, where the agent reasons about your code instead of matching known patterns.
I'd run it on a branch against a mid-sized legacy app and read every diff. A static pass that swaps a JavaScript tooltip for the Popover API or deletes a polyfill is easy to review. It is also easy to merge something that breaks an old browser you still support, so set your Baseline target first.
Numbers from early users
The case studies come from Google's own post, so read them as vendor-reported:
- note inc. pass rate on modern web checks, 180 runs
- 61% → 94%
- MetaMask review comments that led to a fix
- 43 of 64
- MetaMask median time from comment to fix
- 15 min
note inc. ran a 180-run evaluation on its production Next.js codebase. The guides steered agents to native HTML and CSS for scroll animations, tooltips and accordions, and proper LCP image prioritization rose by 28% across 431 merged PRs.
MetaMask runs the guides as a review bot on every pull request to its open source browser extension, with a three-line prompt. In five weeks the bot reviewed about 480 PRs and commented on 64. Engineers fixed 43 findings. Most covered accessibility, HTML semantics and forms, the problems linters miss.
WeTV, Tencent's international streaming service, pairs the guides with Chrome DevTools for agents in CI. The team says its agent resolves 90% of failed performance checks on its own, and resolution time dropped from four hours to 30 minutes.
Pair it with a runtime check
Chrome recommends using the guides together with Chrome DevTools for agents. The guides tell the agent how to break up long tasks; DevTools runs a trace on a throttled CPU and tells the agent whether any task still blocks the main thread past 50 milliseconds, the threshold that feeds INP. The same split applies to animations and memory: the guide shows the pattern, the DevTools tools record the result and track heap growth.
Without the runtime half, your agent writes code that matches a best practice and never checks whether it helped.
This week
- Run
npx modern-web-guidance@latest installin one front-end repo, orupdateif you installed the preview. - Set your browser support target before you run the modernize workflow, and run it on a branch.
- Try the MetaMask setup: a review bot on pull requests with a short prompt. Count how many comments your team acts on after two weeks.
- Open the public eval dashboard and check the guides for the areas your app leans on, such as forms or dialogs, before you trust them in review.
Volodymyr Chornous