v1.0.3 - Documentation Update

📝 Documentation Update

What's Changed:

• Updated all repository links in README to correct GitHub organization

Why this version bump?
NPM freezes the README at publish time, so a patch version is needed to update the package documentation on npmjs.com.

No functional changes - all v1.0.2 features remain identical.

Full Changelog: v1.0.2...v1.0.3

v1.0.2 - Multi-Language Support & Robust Extraction

🌍 Multi-Language Support & Extraction Overhaul

This release brings production-grade multi-language support and a complete rewrite of the extraction engine, making the MCP server significantly more robust and language-independent.

🚀 New Features

Multi-Language Support (6 Languages)

• Full support for DE, EN, NL, ES, FR, IT browser interfaces
• 17 citation selector variants across all languages
• 15 cutoff marker variants for clean content extraction
• Language-independent SVG button detection

4-Stage Completion Detection

• Stage 1 (0-15s): SVG thumbs-up button (100% language-independent)
• Stage 2 (15-30s): Aria-label feedback button fallback
• Stage 3 (30-40s): Multi-language text indicators
• Stage 4 (40s): Smart timeout with partial content extraction

🔧 Technical Improvements

Complete Extraction Rewrite

• Ported proven DOM injection script from Python skill (SERPO method)
• Per-button extraction with smart sidebar change detection
• Fixed citation replacement logic ([CITE-0] → [1][2] based on actual sources)
• Sources now correctly embedded after post-processing
• Code-tag wrapping for citation markers ([CITE-N])

Rendering & Performance

• Removed fixed viewport sizing (prevents oversized browser windows)
• AI Mode availability check with helpful error messages
• Enhanced .gitignore (dist/, node_modules/, browser profiles)

📦 New Files

• src/constants/language-constants.ts - Centralized multi-language constants
• src/utils/completion-detector.ts - Hybrid 4-stage detection logic

🐛 Bug Fixes

• Fixed empty sources array in output
• Fixed citation numbering (now sequential across all sources)
• Fixed markdown source list placement (after post-processing)
• Fixed browser viewport issues causing UI rendering problems

📈 Stats

• 11 files changed: 728 insertions(+), 281 deletions(-)
• MCP server now matches Python skill functionality 1:1

🔗 Related

• Python Skill: google-ai-mode-skill
• NPM Package: npm install google-ai-mode-mcp@1.0.2

v1.0.1 - Codex Linux/WSL Support

What's New

Fixed

• Platform-aware DISPLAY detection: Added Linux-specific X-Server check that prevents errors on Codex (Linux/WSL)
  • Windows/Mac users unaffected (native display servers)
  • Clear error message with xvfb-run configuration instructions
  • Solves "Missing X-Server" error when trying to show browser for CAPTCHA solving

Added

• Gemini installation command in README
• Linux/WSL specific instructions for Codex users
• Detailed troubleshooting guide for X-Server issues

Technical Details

The server now checks process.platform === 'linux' before requiring $DISPLAY environment variable. This ensures:

• Linux/WSL users: Get helpful error with xvfb-run wrapper instructions
• Windows/Mac users: No changes, works as before
• All platforms: Headless mode continues to work without X-Server

Full Changelog: v1.0.0...v1.0.1

v1.0.0 - Initial Release

Google AI Mode MCP Server v1.0.0

First stable release of the Google AI Mode MCP Server.

What This Is

An MCP server that connects any MCP-compatible code agent to Google AI Mode for free, token-efficient web research with grounded sources and inline citations.

Key Features

✅ Free Access - No API keys or subscriptions needed
✅ Multi-Agent Support - Works with Claude Code, Codex, Cursor, Cline, Windsurf, Zed
✅ Query Optimization - Built-in prompt with 5 comprehensive examples
✅ Token Efficient - One synthesized answer instead of reading 5-10 pages
✅ Source Citations - Inline references [1][2][3] with full source list
✅ 3-Layer CAPTCHA Detection - Robust handling with automatic fallback
✅ TypeScript - Full type safety and IntelliSense support

Installation

Claude Code:

claude mcp add google-ai-search npx google-ai-mode-mcp@latest

Codex:

codex mcp add google-ai-search -- npx google-ai-mode-mcp@latest

Other MCP clients (Cursor, Cline, Windsurf, Zed):
Add to your MCP config file:

{
  "mcpServers": {
    "google-ai-search": {
      "command": "npx",
      "args": ["google-ai-mode-mcp@latest"]
    }
  }
}

Quick Start

"Search Google AI Mode for: Next.js 15 App Router best practices"

Your agent will automatically optimize the query and return a clean, cited answer.

npm Package

npm install -g google-ai-mode-mcp

Or use with npx (recommended):

npx google-ai-mode-mcp@latest

Requirements

• Node.js 18.0.0+
• Internet connection

Documentation

• Full README with examples and troubleshooting
• Query optimization guide with 5 examples
• Complete installation instructions for all MCP clients

---

For Claude Code only? Check out the Skill version for simpler installation.