Docs · operator manual

The technical manual for Arcus.

A B2B operator guide: how the audit stack is composed, where AI lives in the pipeline, and what to expect when you wire Arcus into your customer journey.

Where Arcus fits

Overview

Arcus is an agentic audit and delivery platform. It runs ten parallel reviewers against your digital surface area, then channels the highest-leverage findings into a fixed-fee scope of work that VA Arcus engineers ship to production.

The product you're using right now is the audit front-end. Behind it sits a delivery pipeline that turns the audit into a proposal, prototype, and launch — see Pipeline.

What it’s built on

Tech stack

  • FrameworkNext.js 16 (App Router · Turbopack · React 19 · TypeScript strict)
  • StylingTailwind CSS 3 with custom amber/obsidian design tokens
  • AIGoogle Gemini 2.5 Flash via @google/generative-ai with strict JSON schema
  • DatabaseSupabase Postgres (audits, orders) with row-level security
  • EmailResend for transactional lead notifications
  • PWACustom service worker, dynamic manifest, install prompt for iOS & Android
  • HostingEdge-deployable (Vercel, Cloudflare, AWS Amplify all supported)
Always-on phone agents

Voice integrations

The Prestige Dental AI division ships a custom voice stack powered by low-latency speech models, retrieval over the clinic's internal lexicon, and a deterministic state machine for booking flows. The result is a phone agent that answers in under 400 ms, never invents availability, and routes urgent clinical cases to a human within 30 seconds.

Voice tunes itself to your clinical vocabulary, surgeon names, and insurance carriers from a one-page document — no labeled-dataset slog. Call recordings are scored weekly against a rubric you co-design.

Compounding lift

How AI agents increase conversion

A single conversion experiment is a coin flip. Twenty experiments running in parallel, prioritised by audit-validated leverage, is a system. Arcus delivers the system:

  • The audit maps where every visitor leaks — Core Web Vitals, CTA placement, form friction, trust signal density, mobile parity.
  • Each leak becomes a fix queued against an explicit revenue estimate, so the team always works on the next highest-dollar lever.
  • Post-launch, agentic voice and chat capture qualified intent 24/7, lifting the same site without a single new visitor.

Median outcome across our 2025 enterprise cohort: +34% qualified leads in 90 days at flat ad spend.

What plugs in cleanly

Integrations

HubSpot
Salesforce
Pipedrive
Stripe
Payoneer
Twilio
WhatsApp Business
Calendly
Cal.com
Notion
Slack
Linear
Trust by default

Security model

  • Least-privilege keys: public anon key only inserts via row-level policies; service-role keys live exclusively in server-side environment variables.
  • No PII to third parties: domain strings and audit scores go to Gemini, never form submissions, customer records, or identifiable client data.
  • Transport: HTTPS everywhere, HSTS preloaded on production deploys.
  • Data residency: Supabase regions configurable per client.

Read more on the Security page.

Frequently asked

FAQ

How accurate is the audit?
The Gemini-powered scoring is a strong directional baseline. For paid engagements we cross-validate with real crawlers, Lighthouse, and manual QA before the proposal phase.
Can I export audit results?
Yes — every audit row is queryable from your Supabase workspace. CSV/PDF export ships in the v2.7 release.
Do you handle EU customers?
Yes. We can pin Supabase to eu-central-1 and exclude US-hosted LLMs on request.
What if I just want the audit, not the build?
Run as many audits as you want, free. The pipeline is opt-in.