Add audit CTA to /checklist and /mcp pages

docs/checks.md

@@ -482,3 +482,17 @@ First-class agent support. These checks scan your local source code and require
 - Web check implementations: `src/scanner.ts`
 - L5 / `--repo` check definitions: `src/scoring.ts` (`CLI_CHECKS`)
 - L5 / `--repo` check implementations: `src/repo-scanner.ts`
+
+---
+
+## Run a free audit on your site
+
+See exactly which of these 55 checks your site passes — in under 10 seconds.
+
+**[Run a free audit →](https://botvisibility.com)**
+
+Or install the CLI:
+
+```bash
+npx botvisibility <your-url>
+```

docs/mcp.md

@@ -0,0 +1,88 @@
+# MCP integration
+
+BotVisibility scans for [Model Context Protocol](https://modelcontextprotocol.io) (MCP) support as part of its Level 1 and Level 3 checks. Sites that expose MCP tool schemas let AI agents skip manual API discovery and call your tools directly.
+
+---
+
+## Checks
+
+### 1.12 MCP Server (Level 1 — Discoverable)
+
+**What it checks:** A discovery file at `/.well-known/mcp.json` describing your MCP server endpoint and available tools.
+
+**Why it matters:** Without a discovery file, agents must guess whether your site supports MCP and attempt to probe endpoints. With it, they know exactly where your server is and what tools it exposes — zero guessing, no wasted tokens.
+
+**How to pass:**
+
+```json
+// /.well-known/mcp.json
+{
+  "name": "My App MCP Server",
+  "description": "Tools for interacting with My App",
+  "version": "1.0.0",
+  "endpoint": "https://yourapp.com/mcp",
+  "tools": [
+    {
+      "name": "search",
+      "description": "Search for items in the catalog",
+      "inputSchema": {
+        "type": "object",
+        "properties": {
+          "query": { "type": "string", "description": "Search term" }
+        },
+        "required": ["query"]
+      }
+    }
+  ]
+}
+```
+
+Serve this file with `Content-Type: application/json` and no auth required.
+
+---
+
+### 1.18 WebMCP (Level 1 — Discoverable)
+
+**What it checks:** A WebMCP endpoint declaration, enabling browser-accessible MCP tool calls via HTTP.
+
+**Why it matters:** WebMCP allows agents running in browser contexts (web-based AI assistants, browser extensions, Claude.ai) to call your tools without server-side configuration.
+
+**How to pass:** Expose MCP tools over HTTP at a `/mcp` endpoint and declare it in your `/.well-known/mcp.json` discovery file.
+
+---
+
+### 3.7 MCP Tool Quality (Level 3 — Optimized)
+
+**What it checks:** That your MCP tools declared in `/.well-known/mcp.json` have complete definitions: a `name`, a human-readable `description`, and a JSON Schema `inputSchema` for each tool.
+
+**Why it matters:** Incomplete tool definitions force agents to probe parameters by trial and error. Every failed call is wasted tokens and latency. A complete schema lets agents call your tools correctly on the first try.
+
+**How to pass:** Every tool in your discovery file should include:
+
+- `name` — unique identifier for the tool
+- `description` — one or two sentences explaining what the tool does and when to use it
+- `inputSchema` — a JSON Schema object listing all parameters with types and required fields
+
+Partial credit is given when some tools have complete definitions.
+
+---
+
+## Further reading
+
+- [Model Context Protocol spec](https://modelcontextprotocol.io/docs)
+- [MCP tool schema reference](https://modelcontextprotocol.io/docs/concepts/tools)
+- Full check list: [docs/checks.md](checks.md)
+
+---
+
+## Run a free audit on your site
+
+Find out in seconds whether your site passes the MCP checks.
+
+**[Run a free audit →](https://botvisibility.com)**
+
+Or install the CLI:
+
+```bash
+npx botvisibility <your-url>
+```