Archor is an SEO and GEO audit agent that analyzes websites live, scores them against modern search and AI visibility criteria, and turns findings into prioritized action plans and client-ready proposals.
Why this project exists
Most SEO tools produce long, technical reports that business owners cannot act on, and agencies spend hours rewriting those reports into something a client will actually read. Archor collapses that workflow. It audits a site, frames every finding as a business problem rather than a technical one, and hands back a proposal the same session.
What it solves
- Live technical and on-page audits grounded in fetched data, not inference
- GEO and AI search visibility analysis for ChatGPT, Perplexity, and AI Overviews
- Schema, content, Core Web Vitals, local SEO, and backlink coverage
- Prioritized issue lists separated into Critical, High, Medium, and Quick Wins
- Proposal generation that reuses audit context for a consistent client narrative
Product and architecture notes
Archor is split into two pieces. The agent itself runs on Anthropic's Managed Agents platform with Claude Sonnet 4.6 and a custom claude-seo skill that defines the audit commands, scoring model, and consultant persona. The app wrapping it is a React and Vite client with a thin Node and Express backend that proxies requests to the Managed Agents API and keeps credentials server-side. Supabase handles persistence, and the whole thing is deployed on Vercel.
Each client is a first-class object with an audit type, health score, issue counts, stored report, generated proposal, and the session identifiers needed to continue working in the same context on follow-ups and re-audits.
Audit and proposal workflow
Audits stream live through Server-Sent Events so findings appear as the agent works. The agent follows a fixed fetch protocol with rate-limit handling, partial-report fallbacks, and required brand-signal and AI-citation checks for GEO work. Output includes a weighted health score across Technical, Content, On-Page, Schema, Performance, AI Search Readiness, and Images, a what-is-working-well section, and a proposal offer at the end of every audit.
The pipeline view tracks clients through Audit, Action Plan, In Progress, Complete, and Lost stages, with notes and re-audits tied back to the original session so context is never lost between touchpoints.
Current status and direction
Archor is in active use and development, with ongoing work on tighter rate-limit behavior, broader skill coverage, and sharper proposal output. The near-term direction is deeper automation between audit findings and the action plans that follow, so more of the path from prospect to signed engagement happens inside the tool.
My role spans the agent design, skill authoring, app architecture, and the consulting workflow it is built to support.