1. Conduid
  2. Developer Tools
  3. Argus
MCP server · Developer Tools

Argus

All-seeing SEO inspector. Extract meta tags, audit SEO health, get AI-powered insights.

Unclaimed MIT science
59Fair

Scored 3 days ago · breakdown

About Argus

Argus is an MCP server published by pixelesq in the Developer Tools category: all-seeing SEO inspector. Extract meta tags, audit SEO health, get AI-powered insights.

The repository has 6 stars and 1 forks, with the last commit 3 months ago. Recent commit activity is the strongest signal that reported issues will get attention.

Install

uvx
uvx argus-testing
pip
pip install argus-testing

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Tools (4)

compare_seoSide-by-side comparison of 2-5 URLs

Parameters not published.

extract_jsonRaw extraction + audit data as JSON for programmatic use

Parameters not published.

extract_metaExtract all meta tags, OG, Twitter, JSON-LD, headings, links, images

Parameters not published.

seo_auditFull 40+ rule audit with weighted scoring

Parameters not published.

Ask AI

✦

Ask AI about Argus

Powered by Claude · Grounded in docs

I know everything about Argus. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

Static checks derived from the repository 2 days ago. Heuristics on the source, not an audit.

  • ✓README presentThe repository ships documentation.
  • ✓License declaredUsable under a stated license.
  • !Tests presentNo tests found. Behaviour changes may go unnoticed between versions.
  • ✓Dependencies pinnedA lockfile pins dependency versions.
  • ✓No dynamic code executionSource scan found no eval, dynamic Function, or shell execution.
  • !Scoped permissionsDoesn't declare a permission scope. Assume it can do anything its process can.

Releases

v1.0.2v1.0.2
v1.0.1v1.0.1

README


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 — powered by Claude Opus 4.6 for AI insights that rival expensive platforms.


Features

Three-Tab Architecture

Extract — 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 — 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 — 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

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 — runs locally.
3 (Always Available) Static Analysis Rule-based analysis — 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 — Comprehensive prioritized action plan analyzing ALL extracted data (meta, headings, links, images, schema, technical signals) plus audit results
  • Schema Markup Generator — Generates complete, valid JSON-LD ready to paste into your page
  • Technical Fix Generator — Produces exact HTML code fixes for every audit issue
  • Content Gap Analysis — Identifies missing sections based on page type and heading structure
  • Competitor Insights — E-E-A-T signals, featured snippet optimization, and advanced tactics

Why BYOK?

  • Privacy — Your API key never leaves your browser (stored in chrome.storage.local)
  • No backend — Direct browser-to-API calls, no proxy server
  • No cost to us — Users control their own usage
  • Graceful degradation — Works perfectly without any API key (Gemini Nano or static analysis)

API Architecture

Side Panel (React)
    │
    ├─ Has Claude key? → fetch('https://api.anthropic.com/v1/messages')
    │                     with streaming SSE, max_tokens: 2048
    │
    ├─ Gemini Nano available? → Chrome Built-in AI Prompt API (on-device)
    │
    └─ Neither? → Static rule-based analysis (always works)

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 — they work identically in both browser and Node.js runtimes.

Chrome Extension                    MCP Server
─────────────                       ──────────
content.ts (DOM APIs)               fetch() + cheerio (HTML parsing)
        ↓                                   ↓
   PageExtraction ◄───── SAME TYPE ─────► PageExtraction
        ↓                                   ↓
   auditors/engine.ts  ◄── SHARED CODE ──► auditors/engine.ts
        ↓                                   ↓
   AuditReport                          AuditReport (returned via MCP)

Chrome Extension

Quick install: Add to Chrome from the Chrome Web Store.

Or build from source:

  1. Clone and build:
    git clone https://github.com/pixelesq/argus.git
    cd argus
    npm install
    npm run build
    
  2. Open chrome://extensions/, enable Developer mode
  3. Click Load unpacked, select .output/chrome-mv3
  4. Click the Argus icon — the side panel opens

Configure AI (Optional)

Claude Opus 4.6: Click the gear icon in the Insights tab → enter your API key from console.anthropic.com

Gemini Nano: Requires Chrome 138+. Enable chrome://flags/#prompt-api-for-gemini-nano, restart Chrome.


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/
├── entrypoints/              # Chrome extension entry points
│   ├── background.ts         # Service worker (side panel + headers)
│   ├── content.ts            # Content script (DOM extraction)
│   └── sidepanel/            # React side panel UI
│       ├── App.tsx
│       ├── tabs/             # Extract, Audit, Insights
│       └── components/       # UI components
├── lib/                      # Core business logic
│   ├── ai/                   # AI provider system
│   │   ├── claude.ts         # Claude Opus 4.6 API client
│   │   ├── gemini.ts         # Gemini Nano client
│   │   ├── provider.ts       # Unified provider with fallback
│   │   ├── prompts.ts        # System prompts (standard + Opus-enhanced)
│   │   └── settings.ts       # API key management
│   ├── auditors/             # Audit engine (shared with MCP)
│   │   ├── engine.ts         # Orchestration
│   │   ├── scoring.ts        # Weighted scoring algorithm
│   │   └── rules/            # 40+ audit rules across 10 categories
│   └── extractors/           # DOM-based data extraction
├── mcp-server/               # MCP server for Claude Code
│   ├── src/
│   │   ├── index.ts          # Server + tools
│   │   └── extractor.ts      # cheerio-based HTML extraction
│   └── package.json
└── .mcp.json                 # MCP config for Claude Code

Privacy

  • No analytics, no tracking, no data collection
  • The extension makes zero network requests except:
    1. (Optional) HEAD request for HTTP response headers
    2. (Optional) Claude API calls — 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 — see LICENSE for details.


README fetched from the repository 2 days ago. The original is authoritative.

Questions

About Argus

How do I install Argus?

Run uvx argus-testing, then add the server to your MCP client's configuration.

Is Argus safe to use with an AI agent?

Its trust score is 59 out of 100 (fair). It passes 4 of the 6 static security checks run so far; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Argus still maintained?

Yes — the last commit was 3 months ago. The repository has 6 stars and 1 forks.