Skip to content
Beta Free audit

Meet the team

Seven agents. One Web Health Score.

Each named agent owns a specific job. Six read the site in parallel; the seventh reads the others and writes your action list. None claim more than they can deliver, limitations included.

For the discipline-first view (and what activates at each tier) see what we audit.

Agent profiles

  1. The first agent to arrive. Scout reads the site, follows links the way a new visitor would, and decides which five pages best represent the business. Every other agent inherits Scout's page selection.

    Good at

    • Choosing the most representative pages to audit
    • Detecting CMS, frameworks, and tech stack
    • Spotting robots.txt, sitemap.xml, and AI directives
    • Catching dead links in primary navigation

    What it looks for

    • Same-host links matching about, services, pricing, contact, team, work
    • JSON-LD structured data (Organization, LocalBusiness, FAQPage, Person)
    • Title, meta description, canonical, H1 hierarchy
    • robots.txt, sitemap.xml, /llms.txt, /.well-known/llms.txt
    • Pages that 4xx so they never reach the audit

    Honest limitations

    • Five pages only at the free tier. Heavier crawl from £49.
    • Picks pages by URL keyword (English only) and DOM order, not by importance ranking.
    • Sitemap-prioritised crawl is on the roadmap.

  2. Sprint times the site the way a visitor on a mid-range mobile would experience it. HTML weight, request count, hero image size, render-blocking assets. Every second of delay you can't see, Sprint can.

    Good at

    • HTML weight and request count per page
    • Render-blocking scripts and stylesheets
    • Image and font optimisation gaps
    • Mobile-readiness signals

    What it looks for

    • Document size (bytes) and number of inline + external scripts
    • Image count, alt-text coverage, lazy-loading hints
    • Render-blocking CSS in <head>
    • Viewport meta, theme-color, mobile-specific markup
    • HTTP response duration on each crawled page

    Honest limitations

    • Document-only timings at the free tier, not full Lighthouse runs.
    • No real-user metrics (RUM). Lab-only.
    • Full Lighthouse + Core Web Vitals from £49.

  3. Anchor reads the site as a sceptical first-time visitor would. Would I trust them with my data, my card, my enquiry? Trust signals, security headers, named people, third-party reviews, all in scope.

    Good at

    • Spotting missing trust strips above the fold
    • Catching pre-consent third-party trackers
    • Verifying cookie-consent platform presence
    • Recognising real testimonials vs filler quotes
    • Reading Companies House details out of the footer

    What it looks for

    • Above-the-fold trust strip, logos, badges, accreditations
    • Cookie consent platform fingerprints (15+ known CMPs)
    • Security headers: HSTS, CSP, X-Frame-Options, Referrer-Policy
    • Privacy policy and terms reachable in fewer than two clicks
    • Companies House details: registered number, address, directors
    • Named testimonials with company and location
    • External review platform widgets (Trustpilot, Google, Capterra)

    Honest limitations

    • Companies House lookup is fingerprint-only at free tier (footer scan).
    • Full Companies House API verification activates at £149.
    • Cannot verify whether a testimonial actually came from the named person.

  4. Lens is the most scientifically grounded agent. WCAG 2.2 AA, Nielsen heuristics, design discipline, reading age, target-size laws. At paid tiers, a vision-based review of the actual rendered design joins the heuristics.

    Good at

    • Heading hierarchy, alt-text coverage, and language tagging (WCAG)
    • Spotting "no design system" sites by font and weight count
    • Inline-style density as a maintainability and design-discipline signal
    • Reading age and jargon density on hero copy
    • Visual design quality and modernity (paid tiers, vision-based)

    What it looks for

    • H1 presence and uniqueness per page
    • Image alt-text coverage and lang attribute
    • Distinct font families on the page (>3 = visual noise)
    • Distinct font weights (<2 = no hierarchy; >7 = no system)
    • Inline-style density across the DOM
    • Typography hierarchy, spacing, brand polish, modernity (paid: from screenshot)

    Honest limitations

    • Free tier scores design DISCIPLINE only, typography count, weight count, inline-style density. It does not look at the rendered page.
    • Visual quality (modernity, typography polish, layout finesse) requires the paid-tier vision review.
    • Cognitive walkthrough per primary user task is from £149.
    • Cannot evaluate dynamic-content accessibility (focus traps, ARIA live regions in single-page apps).

  5. Compass walks the path from arrival to enquiry. Where does the visitor land, what do they see first, how many fields are in the way before they convert? Baymard CRO benchmarks under the hood.

    Good at

    • Mapping the arrival-to-enquiry journey
    • Counting form fields and friction
    • Spotting weak or absent CTAs above the fold
    • Detecting "PAGE COMING SOON" placeholders in nav

    What it looks for

    • Number of CTAs and their specificity
    • Form field count and required-field ratio
    • Pricing visibility (transparent, hidden, "contact us")
    • Risk-reversal language: guarantees, free trials, money-back
    • Trust strip placement relative to primary CTA

    Honest limitations

    • Heuristic conversion assessment only. Not A/B-test-equivalent.
    • GA4 funnel data and real conversion-rate analysis from £149.
    • Cannot evaluate post-form-submit experience.

  6. Beacon is the agent no incumbent audit tool runs. How well can an AI assistant describe this business? Can Perplexity cite a fact about your service? Is the site structured for LLM consumption, or hidden from it?

    Good at

    • Citability of factual claims about the business
    • Schema markup completeness for LLM consumption
    • Generating a tailored /llms.txt template the owner can deploy
    • Spotting AI-crawler-blocking robots.txt directives

    What it looks for

    • Schema.org types present on About, Services, FAQ pages
    • Author markup and timestamps on content
    • Factual claims (founded year, named clients, locations served) that are extractable
    • GPTBot, ClaudeBot, CCBot, PerplexityBot directives in robots.txt
    • /llms.txt and /.well-known/llms.txt presence

    Honest limitations

    • No live LLM-query simulation in free tier.
    • Live citation testing via Perplexity / Claude / ChatGPT API at £149.
    • AI visibility is an emerging discipline (Q1 2026); benchmarks evolve.

  7. Brief is the last agent to run. After the six pillar agents finish, Brief reads every finding, picks the three to five actions that move the score most, and writes the executive summary.

    Good at

    • Composing the Web Health Score from six pillar inputs
    • Picking the top three to five actions that move the score most
    • Citing the standard behind every action and the cost of waiting
    • Writing the executive summary in plain English

    What it looks for

    • Findings from every pillar agent
    • Cross-pillar patterns (e.g. accessibility + SEO + trust all reference the same image alt-text gap)
    • Severity distribution and commercial impact

    Honest limitations

    • Synthesis only, Brief never reads the site directly.
    • Quality of the summary is bounded by the quality of the pillar findings.
    • Iterative back-and-forth with pillar agents is on the roadmap (single-pass today).

See them work

Watch the seven run on your site.

Submit your URL and watch the dashboard. Each agent narrates what it is reading, in real time, until Brief composes your Web Health Score.

Run my free audit
Run my free audit