🌐 axd-browser-mcp (v1.1.0)

The Sovereign High-Speed Browser & Universal AI Workspace Intelligence MCP Server
100x Faster than Pixel-Based CUA. Zero Vision Token Waste. Universal Multi-Platform Chat Ingestion, Patent Evidence Preservation & Canonical Workspace Reorganization.

🏗️ Multi-Platform & Dual-Engine Architecture

axd-browser-mcp (v1.1.0)
 ├── 🌐 Browser Engine (CDP, Stealth, Bezier Clicks, Fast Extraction) — 8 Tools
 └── 📂 Universal Workspace & Evidence Suite — 4 Tools
      ├── 💬 chats_analyze_and_clean   (Scans, indexes, & dedups Claude & OpenAI sessions)
      ├── 🛡️ ai_evidence_vault_recreate (Ingests Claude, ChatGPT, Perplexity, Gemini into 1 Clean Tree)
      ├── 🌲 projects_reorganize_tree (Audits & structures projects into canonical clean trees)
      └── 🤝 cowork_space_manager    (Syncs, indexes & cleans Claude Desktop Cowork spaces)

Related MCP server: Pilot

⚡ Why axd-browser-mcp?

Traditional AI Computer-Use tools take full 3MB PNG screenshots on every step, upload them to vision LLMs, and guess pixel coordinates.

🚀 Quick Start (Claude Desktop / Cursor / OpenCode)

Add this to your MCP configuration (claude_desktop_config.json or cursor.json):

{
  "mcpServers": {
    "axd-browser-mcp": {
      "command": "npx",
      "args": ["-y", "axd-browser-mcp"]
    }
  }
}

🛠️ Complete MCP Tool Registry (12 Tools)

🌐 High-Speed Browser Automation (8 Tools)

1. browser_navigate: Stealth URL navigation that bypasses bot fingerprinting & Cloudflare.

2. browser_click: Clicks DOM elements using mathematical Bezier curve velocity trajectory.

3. browser_type: Types text with natural human inter-arrival keystroke jitter.

4. browser_press_key: Simulates realistic hardware keypresses (Enter, Tab, Escape, Backspace, etc.).

5. browser_scroll: Smooth wheel scrolling with authentic momentum.

6. browser_screenshot: High-resolution viewport or full-page screen capture.

7. browser_extract: Extracts clean text, page title, URL, and interactive links in <10ms.

8. browser_evaluate: Evaluates arbitrary in-browser JavaScript expressions.

📂 Universal Workspace & Evidence Intelligence Suite (4 Tools)

9. chats_analyze_and_clean:

   • Recursively parses hundreds of .jsonl session files across ~/.claude/projects/ and OpenAI export archives.

   • Extracts conversation dates, token volumes, user topics, and message counts.

   • Identifies broken/stub transcripts (0-byte, error-only) and redundant duplicate threads.

   • Compiles a clean, searchable markdown catalog (MASTER_CHAT_CATALOG.md).

10. ai_evidence_vault_recreate:

    • Universal multi-platform ingestion across Anthropic Claude, OpenAI ChatGPT, Perplexity AI, and Google Gemini.

    • Recreates 1 unified clean-tree directory (AI_EVIDENCE_VAULT/) with normalized Markdown transcripts.

    • Classifies and tags patent IP (Hogan Motor Math, BMS Preservation, Local-First Moat, Legal Records).

    • Compiles MASTER_AI_CATALOG.md, MASTER_DUPLICATE_LEDGER.md, and MASTER_EVIDENCE_REGISTER.md.

11. projects_reorganize_tree:

    • Audits workspace directories across ~/Claude and ~/§00_AXD_Sovereign_empire/.

    • Enforces canonical architecture (01_Product, AI_review, Human_review, Docs, Quarantine).

    • Scans for leaked credentials (.env, OpenAI sk-..., GitHub ghp-..., Slack, AWS keys).

    • Detects junk files (.pyc, __pycache__, .DS_Store) and stages them safely to ~/Quarantine/.

12. cowork_space_manager:

    • Coordinates Claude Desktop's coworkUserFilesPath workspace.

    • Cleans up stale temp files, lock files, and cached index tags.

    • Generates and maintains COWORK_INDEX.md cross-linking all active projects with metadata, stacks, and documentation pointers.

🏛️ AI Evidence Vault Structure

AI_EVIDENCE_VAULT/
 ├── 00_MASTER_INDEX/
 │    ├── MASTER_AI_CATALOG.md          <-- Universal Master Index across ALL platforms
 │    ├── MASTER_DUPLICATE_LEDGER.md    <-- Duplicate thread clusters & consolidation map
 │    └── MASTER_EVIDENCE_REGISTER.md   <-- All identified Patents, Breakthroughs & Legal Records
 ├── 01_Anthropic_Claude/
 │    ├── Web_Conversations/             <-- Normalized Markdown transcripts (.md)
 │    └── CLI_and_Desktop_Sessions/      <-- Normalized Markdown transcripts (.md)
 ├── 02_OpenAI_ChatGPT/
 │    ├── Personal_Diego/                <-- Normalized Markdown transcripts (.md)
 │    ├── Hotmail_Archive/               <-- Normalized Markdown transcripts (.md)
 │    └── Gmail_Archive/                 <-- Normalized Markdown transcripts (.md)
 ├── 03_Perplexity_AI/                   <-- Research threads with citation sources
 ├── 04_Google_Gemini_Antigravity/       <-- Gemini & Antigravity transcripts
 ├── 05_Forensic_Evidence_Extracts/      <-- Curated dossiers of Patent & Legal evidence
 └── 06_Quarantine_Stubs/                <-- Isolated 0-byte/empty transcripts

🧪 Testing

# Run unit & integration tests
npm test

# Run live E2E MCP JSON-RPC protocol test
npm run test:e2e

💼 Commercial & Pro Tier

• Free Tier: 100 autonomous browser actions & workspace scans per month.

• Pro Tier ($29/mo): Unlimited actions, multi-tab orchestration, and residential proxy routing.

Author: Diego Nanini
Inquiries: diegonanini23@gmail.com