Argus
All-seeing SEO inspector. Extract meta tags, audit SEO health, get AI-powered insights.
Installation
npx argusAsk AI about Argus
Powered by Claude Β· Grounded in docs
I know everything about Argus. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Reviews
Documentation
Argus
The all-seeing, AI-powered SEO inspector.
A free Chrome extension + MCP server that brings enterprise-level SEO auditing to everyone β powered by Claude Opus 4.6.
getargus.app
Website Β· Chrome Web Store Β· npm Β· Features Β· Install Β· Opus 4.6 Integration Β· MCP Server Β· Architecture
The Problem
Enterprise SEO tools cost $100-300/month. SEO expertise is locked behind years of specialist knowledge. Most free tools only scratch the surface.
Argus breaks these barriers. It's a completely free, open-source Chrome extension that gives you instant, deep SEO analysis on any webpage β powered by Claude Opus 4.6 for AI insights that rival expensive platforms.
Features
Three-Tab Architecture
Extract β Every SEO tag at a glance
- Title, meta description, canonical, robots (with character count badges)
- Open Graph & Twitter Card tags with live social preview cards
- JSON-LD structured data with syntax highlighting and validation
- Heading hierarchy as a visual tree
- Link analysis (internal/external/nofollow counts)
- Image audit (missing alt text, dimensions, lazy loading, modern formats)
- One-click copy and full JSON/text export
Audit β 40+ rule-based SEO checks
- Weighted scoring (0-100) across 10 categories
- Categories: title, description, headings, images, links, technical, structured data, social, content, performance
- Color-coded severity (critical/warning/info/pass) with fix guidance
- Real Web Vitals: LCP, INP, CLS, TTFB via PerformanceObserver API
Insights β AI-powered deep analysis
- Three-tier AI provider system with automatic fallback
- Automatic page classification, content quality, and SEO strategy analysis
- On-demand: rewrite meta descriptions, improve titles, suggest FAQ schema, generate schema markup
- Opus 4.6 exclusive: SEO strategy brief, technical fixes, content gap analysis, competitor insights
Opus 4.6 Integration
Argus uses Claude Opus 4.6 as its primary AI engine via a BYOK (Bring Your Own Key) architecture:
Three-Tier AI Provider System
| Tier | Provider | How It Works |
|---|---|---|
| 1 (Primary) | Claude Opus 4.6 | User provides their API key (stored locally, never synced). Direct fetch() from the side panel to api.anthropic.com with streaming SSE. |
| 2 (Fallback) | Gemini Nano | Chrome's on-device AI model. No API key, no network β runs locally. |
| 3 (Always Available) | Static Analysis | Rule-based analysis β works offline with zero dependencies. |
Opus 4.6 Exclusive Features
When Claude is configured, Argus unlocks advanced analysis that smaller models can't handle:
- SEO Strategy Brief β Comprehensive prioritized action plan analyzing ALL extracted data (meta, headings, links, images, schema, technical signals) plus audit results
- Schema Markup Generator β Generates complete, valid JSON-LD ready to paste into your page
- Technical Fix Generator β Produces exact HTML code fixes for every audit issue
- Content Gap Analysis β Identifies missing sections based on page type and heading structure
- Competitor Insights β E-E-A-T signals, featured snippet optimization, and advanced tactics
Why BYOK?
- Privacy β Your API key never leaves your browser (stored in
chrome.storage.local) - No backend β Direct browser-to-API calls, no proxy server
- No cost to us β Users control their own usage
- Graceful degradation β Works perfectly without any API key (Gemini Nano or static analysis)
API Architecture
Side Panel (React)
β
ββ Has Claude key? β fetch('https://api.anthropic.com/v1/messages')
β with streaming SSE, max_tokens: 2048
β
ββ Gemini Nano available? β Chrome Built-in AI Prompt API (on-device)
β
ββ Neither? β Static rule-based analysis (always works)
MCP Server β SEO Auditing in Your Terminal
Argus includes a Model Context Protocol (MCP) server that brings SEO auditing directly into Claude Code. Developers can audit any webpage's SEO without leaving their terminal.
Tools
| Tool | Description |
|---|---|
seo_audit | Full 40+ rule audit with weighted scoring |
extract_meta | Extract all meta tags, OG, Twitter, JSON-LD, headings, links, images |
compare_seo | Side-by-side comparison of 2-5 URLs |
extract_json | Raw extraction + audit data as JSON for programmatic use |
Setup
One-line install (recommended):
claude mcp add argus-seo -- npx -y @pixelesq/argus-mcp
Or build from source:
# Build the MCP server
cd mcp-server && npm install && npm run build
# Add to Claude Code
claude mcp add argus-seo -- node mcp-server/build/index.js
Or drop a .mcp.json in any project:
{
"mcpServers": {
"argus-seo": {
"command": "node",
"args": ["path/to/argus/mcp-server/build/index.js"]
}
}
}
Shared Audit Engine
The MCP server shares the same audit engine as the Chrome extension. The audit rules are pure functions that take a PageExtraction data structure β they work identically in both browser and Node.js runtimes.
Chrome Extension MCP Server
βββββββββββββ ββββββββββ
content.ts (DOM APIs) fetch() + cheerio (HTML parsing)
β β
PageExtraction ββββββ SAME TYPE ββββββΊ PageExtraction
β β
auditors/engine.ts βββ SHARED CODE βββΊ auditors/engine.ts
β β
AuditReport AuditReport (returned via MCP)
Install
Chrome Extension
Quick install: Add to Chrome from the Chrome Web Store.
Or build from source:
- Clone and build:
git clone https://github.com/pixelesq/argus.git cd argus npm install npm run build - Open
chrome://extensions/, enable Developer mode - Click Load unpacked, select
.output/chrome-mv3 - Click the Argus icon β the side panel opens
Configure AI (Optional)
Claude Opus 4.6: Click the gear icon in the Insights tab β enter your API key from console.anthropic.com
Gemini Nano: Requires Chrome 138+. Enable chrome://flags/#prompt-api-for-gemini-nano, restart Chrome.
Architecture
Tech Stack
| Layer | Technology |
|---|---|
| Extension Framework | WXT (Vite-based) |
| UI | React 19 + TailwindCSS 4 |
| Language | TypeScript (strict) |
| Manifest | Chrome Extension Manifest V3 |
| AI (Cloud) | Claude Opus 4.6 via Messages API |
| AI (On-Device) | Chrome Built-in AI (Gemini Nano) |
| MCP Server | @modelcontextprotocol/sdk + cheerio |
| Icons | Lucide React |
Project Structure
argus/
βββ entrypoints/ # Chrome extension entry points
β βββ background.ts # Service worker (side panel + headers)
β βββ content.ts # Content script (DOM extraction)
β βββ sidepanel/ # React side panel UI
β βββ App.tsx
β βββ tabs/ # Extract, Audit, Insights
β βββ components/ # UI components
βββ lib/ # Core business logic
β βββ ai/ # AI provider system
β β βββ claude.ts # Claude Opus 4.6 API client
β β βββ gemini.ts # Gemini Nano client
β β βββ provider.ts # Unified provider with fallback
β β βββ prompts.ts # System prompts (standard + Opus-enhanced)
β β βββ settings.ts # API key management
β βββ auditors/ # Audit engine (shared with MCP)
β β βββ engine.ts # Orchestration
β β βββ scoring.ts # Weighted scoring algorithm
β β βββ rules/ # 40+ audit rules across 10 categories
β βββ extractors/ # DOM-based data extraction
βββ mcp-server/ # MCP server for Claude Code
β βββ src/
β β βββ index.ts # Server + tools
β β βββ extractor.ts # cheerio-based HTML extraction
β βββ package.json
βββ .mcp.json # MCP config for Claude Code
Privacy
- No analytics, no tracking, no data collection
- The extension makes zero network requests except:
- (Optional) HEAD request for HTTP response headers
- (Optional) Claude API calls β only when user provides their own key, direct to Anthropic
- Gemini Nano runs entirely on-device
- API keys stored in
chrome.storage.local(device-local, never synced)
Development
# Extension dev server (hot reload)
npm run dev
# Production build
npm run build
# MCP server
cd mcp-server && npm install && npm run build
License
MIT License β see LICENSE for details.
Built for the Built with Opus 4.6: a Claude Code Hackathon
