Skip to content

About

Lighthouse-style Chrome extension for auditing websites for AI agent compatibility.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

Repository files navigation

GetAgentReady

A Chrome extension that audits any website for AI agent compatibility and returns a 0–100 score — like Lighthouse, but for agents.

The web was built for humans. As AI agents (Claude, GPT-based tools, LangChain pipelines, browser-use, etc.) increasingly need to interact with websites programmatically, most sites are entirely opaque to them. GetAgentReady gives developers and site owners a one-click audit that surfaces exactly what to fix.


Install (development)

  1. Clone or download this repo
  2. Open chrome://extensions
  3. Enable Developer mode (top right)
  4. Click Load unpacked → select the project folder
  5. Navigate to any page and click the GetAgentReady toolbar icon → Run Audit

The DevTools panel is available under the GetAgentReady tab in Chrome DevTools (⌘⌥I / F12).


How it works

Click Run Audit in the popup or DevTools panel. The extension:

  1. Injects a content script that collects DOM data (headings, JSON-LD, forms, links, images, dates, etc.)
  2. Fires all network checks in parallel from the service worker (no CORS restrictions)
  3. Runs 32 checks across 6 domains
  4. Computes a weighted composite score 0–100
  5. Caches the result for 10 minutes; shows a score delta on re-run

Everything runs client-side. No data leaves your browser.


Scoring

domain_score = Σ(weight × outcome_value) / Σ(weight) × 100
composite    = Σ(domain_score × domain_weight)

PASS = 1.0 · WARN = 0.5 · FAIL = 0.0 · N/A excluded from denominator
Band Range Colour
Excellent 90–100 Green
Good 70–89 Light green
Needs Work 50–69 Amber
Poor 25–49 Orange
Critical 0–24 Red

Domain weights

Domain Weight
Content Parseability 30%
Discoverability 25%
API & Integration 20%
Interactability 12%
Agent Policy 8%
Authentication 5%

The 32 checks

Discoverability

ID Check Weight What it tests
D1 llms.txt 5 /llms.txt exists, is non-empty, and follows the llmstxt.org spec (headings, block quotes)
D2 robots.txt AI directives 5 robots.txt contains explicit User-agent blocks for AI crawlers (GPTBot, ClaudeBot, PerplexityBot, etc.)
D3 sitemap.xml 4 /sitemap.xml accessible, contains <url> entries with <lastmod> dates
D4 OpenAPI spec 4 Machine-readable API spec discoverable via <link rel="describedby">, /.well-known/openapi.json, or common paths
D5 MCP server 3 Model Context Protocol manifest present at /.well-known/mcp

Content Parseability

ID Check Weight What it tests
P1 SSR content 5 Static HTML word count ≥ 80% of rendered (detects CSR-only shells)
P2 Heading hierarchy 5 Single <h1>, no skipped levels — primary navigation signal for agents
P3 JSON-LD structured data 5 At least one valid Schema.org JSON-LD block present
P4 Meta tag completeness 4 description, og:title, og:description, and <link rel="canonical"> all present
P5 ARIA landmarks 4 <main>, <nav>, <header>, <footer> (or role equivalents) present
P6 Alt text coverage 3 ≥ 80% of non-decorative images have non-empty alt attributes
P7 Link descriptiveness 3 No links with vague text ("click here", "read more", "here", etc.)
P8 Markdown content 4 Markdown version discoverable via link tag, Accept: text/markdown, .md URL, or /llms-full.txt
P9 Last updated date 3 Visible date labeled "Last updated / Updated / Modified" — or article:modified_time meta tag
P10 JSON-LD date fields 3 datePublished and/or dateModified present in Schema.org JSON-LD
P11 Copyright notice 2 Copyright notice with year (© YYYY) present, ideally in <footer>

API & Integration

ID Check Weight What it tests
A1 OpenAPI spec validity 4 Discovered spec is valid OpenAPI 3.x with paths and request/response schemas
A2 Feed availability 3 RSS or Atom feed discoverable via <link rel="alternate"> or common paths
A3 GraphQL introspection 1 GraphQL endpoint responds to { __typename } introspection

Authentication

ID Check Weight What it tests
AU1 OAuth/OIDC well-known 2 /.well-known/openid-configuration or /.well-known/oauth-authorization-server present
AU2 Auth method disclosure 2 WWW-Authenticate header on protected resources, or auth docs linked from page
AU3 API key documentation 2 API authentication documentation reachable within one hop

Agent Policy

ID Check Weight What it tests
PL1 AI bot policy direction 5 robots.txt explicitly allows or blocks AI crawlers (informational — blocking is a legitimate choice)
PL2 Rate limiting headers 3 X-RateLimit-Limit, X-RateLimit-Remaining, or Retry-After headers present
PL3 Content freshness signals 2 Last-Modified and/or ETag response headers enable conditional GETs
PL4 CORS policy 2 Access-Control-Allow-Origin headers present on API endpoints
PL5 ToS AI clause 1 Terms of Service contain language about AI / automated access
PL6 AI exclusion signals 2 noai / noimageai directives in <meta name="robots"> or X-Robots-Tag (informational)

Interactability

ID Check Weight What it tests
I1 Form field labeling 5 All interactive form fields have <label for>, aria-label, or aria-labelledby
I2 Stable element IDs 1 No duplicate id attributes in the DOM
I3 Form action discoverability 3 Forms have explicit action + method and a native submit button
I4 Focus management 1 No tabindex > 0 abuse

UI

Toolbar popup

  • Score ring (0–100, colour-coded by band)
  • Domain breakdown bars
  • Issue counts (failures / warnings / passed)
  • Top 3 failures with one-line fix hints
  • Export JSON button
  • View all checks link → opens details page

DevTools panel (GetAgentReady tab)

  • Full score overview + domain bars
  • Filter bar: All · Failures · Warnings · Passed · N/A
  • Domain accordion sections — each check row expands to show description, fix, and reference link
  • Weight indicator (5 dots) per check
  • Score delta on re-run (e.g. +8 or -3)
  • Export JSON and Export HTML buttons

HTML report export

A standalone self-contained HTML file with:

  • Full audit results with interactive filter buttons and click-to-expand check detail
  • Schema.org Report JSON-LD in <head> with datePublished and dateModified — the report passes its own checks
  • Copyright footer and last-updated timestamp

Architecture

manifest.json                   MV3, ES module service worker

background/
  service-worker.js             Orchestrates audits · network checks · badge · cache

content/
  content-script.js             Self-contained DOM data collector
                                (headings, JSON-LD, meta, forms, images, links,
                                 landmarks, dates, copyright, interactivity)

audit/
  engine.js                     Runs all 32 checks; returns check result array
  scorer.js                     Weighted domain + composite score computation
  dom-checks.js                 Pure functions over domData (no network)
  network-checks.js             All fetch-based checks (service worker context)
  export.js                     Generates standalone HTML report

shared/
  constants.js                  CHECK_DEFS · DOMAIN_WEIGHTS · OUTCOME · score bands
  utils.js                      fetchWithTimeout · parseRobotsTxt · findAiBotRules

popup/
  popup.html / popup.js / popup.css

devtools/
  devtools.html / devtools.js   Registers the DevTools panel
  panel.html / panel.js / panel.css

options/
  options.html / options.js / options.css

Execution split

Content script — runs in the page, reads the DOM synchronously: headings, JSON-LD, meta tags, ARIA landmarks, forms, images, links, feed links, <time> elements, copyright text, markdown link tags, meta robots.

Service worker — no CORS restrictions; fires all network fetches in parallel: robots.txt, llms.txt, sitemap.xml, OpenAPI spec, MCP well-known, SSR comparison fetch, feed reachability, GraphQL probe, OAuth well-known, WWW-Authenticate headers, rate limit headers, freshness headers, CORS preflight, ToS fetch, markdown content negotiation, X-Robots-Tag.

Popup / Panel — request audit via message, read result from chrome.storage.session, render UI, trigger exports.

Data flow

User clicks "Run Audit"
  → popup / panel sends { type: 'RUN_AUDIT', tabId, url }
    → service worker sends { type: 'GET_DOM_DATA' } to content script
    → service worker fires all network checks in parallel
    → engine.js aggregates domData + networkData → 32 check results
    → scorer.js computes domain scores + composite
    → result stored in chrome.storage.session (10 min TTL)
    → toolbar badge updated (score + band colour)
  → popup / panel re-renders with result + optional score delta

Caching & score delta

Results are cached in chrome.storage.session keyed by audit_{tabId}. Before overwriting, the previous composite score is captured and stored as scores.previousScore so the UI can display the delta on re-run. The cache is cleared when the tab navigates.

Permissions

Permission Why
activeTab Access current tab ID and URL
storage Cache audit results + settings
scripting Inject content script on demand
<all_urls> Service worker fetches to any origin (robots.txt, llms.txt, etc.)

Settings

Setting Default Description
Auto-run on page load Off Run audit automatically on navigation
Show injected badge Off Floating score badge on audited pages
Score display mode Numeric Numeric / Letter grade / Colour only

Verification reference sites

Site Expected behaviour
stripe.com/docs High score — SSR, structured data, OpenAPI
anthropic.com Check for llms.txt (D1)
CSR-only React app Low P1 score (SSR fail)
Site without robots.txt D2 fail
Own fixture page Deterministic all-PASS run

Determinism check: running the same audit twice on the same page state must produce identical scores.


Roadmap

Phase 3 (next)

  • Scoring customisation (adjust domain weights)
  • Check filtering by agent type (browser agent vs API agent)
  • Batch mode — sample up to 10 pages from sitemap
  • CLI companion for CI/CD use

Phase 4

  • GitHub README badge (GetAgentReady Score: 82)
  • llmstxt.org validator integration
  • Community aggregate scores by site category

About

Lighthouse-style Chrome extension for auditing websites for AI agent compatibility.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages