v1.6.0 — Site Cloner, Crawler, Template Builder

What's new in v1.6.0

Three major features, all working with OpenRouter's free models:

Site Cloner — MCP_Site_Cloner (single + multi-page)

• POST /wp-json/mcp/v1/clone to clone any public URL into an Elementor page
• Multi-page mode discovers nav + probes /about, /contact, /blog to build a full site

Template Builder — MCP_Template_Builder (no URL needed)

• Build a complete Elementor site from a brand brief
• 7 design systems: Modern SaaS, Warm Editorial, Bold Studio, Calm Spa, Restaurant Warm, Tourism Vivid, Persian Traditional
• 7 role-based section builders: header, hero, features, about, testimonials, cta, footer
• Each section is generated as a separate AI call with a focused role prompt for reliability

Site Crawler — MCP_Site_Crawler

• Walks nav links, resolves relative paths, classifies by URL+anchor into roles
• Path probing: HEAD-probes common paths to fill in missing nav links
• SSRF guards: blocks localhost, 127.0.0.1, private/reserved IP ranges

Under the hood

• New MCP_Site_Fetcher with cURL, robots.txt, asset extraction
• New MCP_DOM_Analyzer for sections, columns, palette, typography
• New MCP_AI_Translator with HTML condensation and JSON parse fallback
• New bracket-aware JSON extractor handles '[text] [{...}]' and other noise prefixes
• Default model: nvidia/nemotron-3-super-120b-a12b:free
  (Llama 3.3 70B no longer free on OpenRouter)

Tests

• 99 → 153, all passing
• No PHPUnit dependency — standalone runner in tests/test-runner.php

Stats

• 32 → 39 PHP files
• 4174 → 6225 LOC

See README.md for full feature list and readme.txt for WordPress changelog.