VigIA – AI Visibility, Analytics & Control

The AI Visibility Score is a 100-point rating system that measures how well your site is prepared for AI crawlers and AI-powered search. It checks 20 signals across 5 categories: access and discovery, structured data, content structure, AI interaction, and performance. You get a letter grade (A+ to F) and specific recommendations to improve.

No. VigIA adds minimal overhead by checking the User-Agent string on each request. The check is very fast and only writes to the database when an AI crawler is detected.

Robots.txt is advisory – crawlers should respect it but may choose to ignore it. PHP blocking returns a 403 Forbidden response, effectively preventing access regardless of whether the crawler respects robots.txt.

Blocking AI training crawlers (like GPTBot or ClaudeBot) will not affect traditional search engine rankings. However, blocking AI search crawlers might affect how your content appears in AI-powered search results.

VigIA only believes proxy headers from a proxy it can trust: Cloudflare, or a proxy on your own network. Anybody can send those headers, so trusting them from anywhere would let a blocked visitor in.

If a load balancer with a public address, or a CDN other than Cloudflare, sits in front of your site, declare it from a plugin or a must-use plugin:

add_filter( 'vigia_trusted_proxies', function () { return array( '203.0.113.10', '198.51.100.0/24' ); } );

List every proxy in front of PHP, including a local one.

Markdown for Agents is a standard for serving web content as clean markdown to AI agents. Instead of processing full HTML, agents receive lightweight markdown with structured metadata. VigIA supports both dedicated .md URLs and Accept: text/markdown content negotiation. Enable it in VigIA > Extras > Markdown for Agents.

This feature also covers taxonomy archives (categories, tags, WooCommerce product categories, custom taxonomies) and WooCommerce products. Term archives include the term description, the list of child terms in hierarchical taxonomies and an excerpt of the latest entries; product .md endpoints embed schema-like data (price, sale price, SKU, stock status, rating) directly in the YAML frontmatter.

VigIA classifies each crawler hit into one of these buckets, indexed in the database so filters and CSV exports are instant: Home (the / path), Post, Page, Product, any other public custom post type, Category archive, Tag archive, Date/Author archive, Feed, Sitemap, REST API, File (PDFs, images, downloads), Admin/login attempt ( /wp-admin , wp-login.php — useful to spot bots probing the admin), WordPress system (admin-ajax, xmlrpc, wp-cron, wp-comments-post), 404 Not found, and Other.

Yes! In the main analytics page, scroll down to “Custom crawlers” and add your own User-Agent patterns to track.

All data is stored in your WordPress database in a custom table ( wp_vigia_visits ). No data leaves your server.

The llms.txt file is a standard for helping AI systems understand your website’s content and structure. It provides a machine-readable overview of your site that AI can use to better represent your content. Learn more at llmstxt.org.

Yes. On a network each site keeps its llms.txt and llms-full.txt in its own uploads folder and WordPress serves them at that site’s address, be it a subdomain, a subdirectory or a mapped domain. Generate them from VigIA > Extras > LLMs on each site. If the main site already has physical files at the network root, set its Delivery to Served by WordPress first: the web server answers a file in that folder on every domain of the network, ahead of each site’s own.

On a single site the files are physical files at the site root unless you choose Served by WordPress under Delivery, which is the option to pick when that folder is not writable.

The file is fine. Your web server sends it without saying it is UTF-8, so the browser guesses another encoding.

• Set Delivery to Served by WordPress and generate again. WordPress sends the files as UTF-8.
• Or ask your host to add a UTF-8 charset to .txt files.

VigIA checks this each time you generate the files and warns you in the Delivery section.

VigIA supports automatic noindex detection from: Yoast SEO, Rank Math, All in One SEO, SEOPress, The SEO Framework, and Native SEO NoIndexer.

The Abilities API is a new feature in WordPress 6.9 that allows plugins to expose their functionality in a standardized way. This enables AI agents, automation tools, and external systems to discover and use plugin features programmatically. VigIA implements 9 abilities for analytics, blocking, and robots.txt management.

Yes: the free MCP Adapter plugin, the official bridge between WordPress and MCP clients.

• The MCP server is off by default. Turn it on in VigIA > Extras > MCP .
• The same tab offers to install MCP Adapter when it is missing.
• The rest of VigIA works without it.

Save the JSON block from Quick Connect as claude_desktop_config.json in your user Library (this is not the system Library at the root of the disk):

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

On macOS, the easiest way to reach the folder is to open Finder, press ⌘ Shift G, paste ~/Library/Application Support/Claude/ and hit Enter. On Windows, press Win+R and run %APPDATA%\Claude .

Important: Claude Desktop only speaks stdio to local processes , so the snippet does not connect directly to VigIA over HTTP. Instead it launches a small bridge package ( mcp-remote ) via npx that proxies the connection. This means you need Node.js installed on the machine. The first run downloads mcp-remote automatically; subsequent runs use the npm cache.

If you do not want to install Node.js, connect from Claude Code or Cursor instead — both speak HTTP MCP natively and do not need a bridge.

Restart Claude Desktop after saving the file. If the app boots with default preferences, the JSON is malformed — review the file or restore your backup. If Claude Desktop says the entry is “not a valid MCP server configuration”, npx is not in its PATH; check that Node.js is installed and accessible to GUI apps.

If the file already exists with other content, see the next question.

If your claude_desktop_config.json or ~/.cursor/mcp.json already exists, do not paste the full Quick Connect block on top of it. Pasting on top discards everything else (preferences, other MCP servers) and the app will start with defaults.

Always make a backup of the file first. Then open it with any text editor that preserves JSON.

There are two scenarios.

Scenario 1 — the file has content but no mcpServers block yet.

This is common when you have used Claude Desktop before but never configured MCP servers. The file might look like this:

{ "preferences": { "menuBarEnabled": false, "...": "..." } }

Add mcpServers as a sibling property of preferences , separated by a comma. The result should be:

{ "preferences": { "menuBarEnabled": false, "...": "..." }, "mcpServers": { "vigia-your-site-example": { "command": "npx", "args": [ "-y", "mcp-remote", "https://your-site.example/wp-json/vigia/v1/mcp", "--header", "Authorization: Basic BASE64_OF_USER_AND_APP_PASSWORD" ] } } }

The order of preferences and mcpServers is not important, but the comma between them is required. Forgetting the comma makes the JSON invalid and Claude Desktop will start with default preferences.

Scenario 2 — the file already has mcpServers with other servers.

Add the VigIA entry inside the existing mcpServers object, separated from other entries by a comma:

"mcpServers": { "other-server": { "...": "..." }, "vigia-your-site-example": { "command": "npx", "args": [ "-y", "mcp-remote", "https://your-site.example/wp-json/vigia/v1/mcp", "--header", "Authorization: Basic BASE64_OF_USER_AND_APP_PASSWORD" ] } }

For Cursor the entry is different — Cursor speaks HTTP MCP natively, so its block uses type , url and headers directly inside the server entry instead of the bridge command. The Quick Connect panel renders the right format for each client.

Most MCP clients accept HTTP transport with a custom Authorization header. The Quick Connect panel exposes the two raw values you need — the server URL and the Authorization header — so you can drop them into whatever configuration format your client expects.

Browser-only assistants without an MCP client (AI Studio, ChatGPT web) cannot connect. They need a desktop or CLI client that speaks MCP over HTTP.

JSON-LD (JavaScript Object Notation for Linked Data) is structured data that helps search engines and AI systems understand your site identity and content. VigIA generates two types of JSON-LD: Site Identity (WebSite + Organization/Person schema with social profiles) and AI Discovery (ReadAction pointers to your llms.txt and Markdown for Agents endpoints). This makes your AI-ready content discoverable through structured signals. Enable it in VigIA > Extras > JSON-LD.