Back to Work
Open Source

Raptor

Autonomous SEO audit engine with 114+ analyzers, health scoring, AI-driven analysis, and Cloudflare Browser Rendering. Built in Rust for speed and reliability.

View Live SiteGet Your Free Audit
Raptor

Duration

Ongoing

Team Size

1 developer

Industry

SEO & Developer Tools

Client

Open Source

Project Results

114+

SEO analyzers across technical, on-page, and structured data

<10s

Full site audit on most small business websites

4 formats

JSON, HTML, CSV, and SVG report output

The Challenge

Most SEO audit tools are slow, expensive, and treat their analysis like a black box. They spit out generic recommendations that don’t account for context, miss JavaScript-rendered content entirely, and charge hundreds per month for the privilege. I needed a tool that could crawl a site fast, render SPAs properly, and produce structured data that an AI could actually reason about, not just flag keywords.

Our Solution

I built Raptor from scratch in Rust. It’s an async concurrent crawler with 114+ SEO analyzers covering everything from title tags and schema markup to security headers and content readability. It integrates with Cloudflare Browser Rendering to handle JavaScript-heavy sites that traditional crawlers miss entirely. The output is structured JSON designed to feed directly into Claude for intelligent analysis, or exposed as an MCP server for agentic workflows where the AI crawls, analyzes, and acts autonomously.

Key Features

Async Concurrent Crawling - Configurable concurrency with semaphore-bounded parallel requests, robots.txt compliance, and sitemap discovery
114+ SEO Analyzers - Titles, meta descriptions, headings, images, links, structured data, security headers, canonicals, hreflang, content quality scoring
Health Scoring Engine - 100-point scale with per-issue weighted deductions and category caps. Grades from A to F
Cloudflare Browser Rendering - Full JavaScript rendering for SPAs. Static vs. rendered comparison reveals content invisible to traditional crawlers
AI-Powered Analysis - Structured JSON output designed for Claude API consumption. Turns raw audit data into prioritized, context-aware fix plans
MCP Server Integration - Expose Raptor as tools Claude can invoke autonomously for competitive audits, regression monitoring, and fix verification
Self-Contained Reports - Single-file HTML reports with all CSS, JS, and SVG inlined. Works offline, zero external dependencies
Entity Extraction - AI-powered extraction of business data (name, phone, address, services, hours) using Workers AI

Technologies Used

Rust
Tokio
Cloudflare Workers
Claude API
MCP
Python
spaCy
Minijinja

Technical Implementation

frontend

Self-contained HTML reports built with Minijinja templates compiled directly into the binary. Light theme CSS, interactive tabs for Overview, Issues, Pages, and Structure views. SVG site structure visualization. A mascot whose mood changes based on the audit score.

backend

Core engine written in Rust using Tokio for async I/O. BFS crawl queue with configurable depth and page limits. Event-driven architecture using CrawlEvent channels that decouple the engine from any UI, making it embeddable as a library.

integrations

Cloudflare Browser Rendering for JavaScript-heavy sites. Claude API integration for intelligent analysis. MCP server for agentic workflows. Python companion tools for NER (spaCy), internal link graph visualization, and historical SERP tracking via Wayback Machine.

performance

Rust’s zero-cost abstractions and Tokio’s async runtime deliver high-throughput crawling. Responsible by default: honors robots.txt, polite delays, exponential backoff, rate limiting. Library-first architecture means zero CLI coupling.