MCP Server
Connect Claude Desktop and any MCP-compatible AI agent to ByteWaveNetwork's SEO and performance tools. Run full site audits, trace redirects, check page speed, and more — directly from your AI assistant, no browser required.
What is MCP?
Model Context Protocol (MCP) is an open standard for connecting AI assistants to external tools and data sources. Instead of copying URLs into a chat window, your AI agent can call ByteWaveNetwork tools directly and receive structured results it can reason about.
ByteWaveNetwork implements MCP 2024-11-05 over HTTP POST (stateless JSON-RPC 2.0). No persistent connection is required — each request is independent.
Claude Desktop Setup
Add the following to your claude_desktop_config.json (found under ~/Library/Application Support/Claude/ on macOS or %APPDATA%\Claude\ on Windows):
Restart Claude Desktop after saving. You should see "ByteWaveNetwork" appear in the MCP servers list. You can now ask Claude things like:
- "Check example.com for broken links"
- "Audit the SEO health of mysite.com"
- "Trace the redirect chain for this URL"
- "How fast does example.com load?"
Available Tools
Eight tools are available. Three tools start async BFS crawls and return a scan_id for polling; five tools return results synchronously.
| Tool name | Mode | Description |
|---|---|---|
| link_checker | async | BFS crawl for broken links, redirects, slow pages, canonical issues, and noindex pages. Required: url. Optional: maxPages (default 50), concurrency (default 5). |
| seo_analyzer | async | BFS crawl scoring each page 0–100 across 22 SEO checks (title, description, headings, canonical, OG, JSON-LD, noindex). Required: url. Optional: depth (default 2). |
| sitemap_validator | async | Fetch and validate an XML sitemap — checks every URL for broken links, noindex, and canonical mismatches. Required: url. Optional: maxUrls (default 200). |
| redirect_tracer | sync | Follow the full redirect chain with per-hop status codes, timing, headers, and loop detection. Required: url. Optional: maxHops (default 20). |
| page_speed_inspector | sync | Measure TTFB, DNS, TCP, TLS timing, HTTP version, compression, cache policy, and 12+ HTML performance checks. Required: url. |
| schema_tester | sync | Extract and validate JSON-LD, Microdata, and RDFa against 18 Schema.org types. Required: url. |
| page_seo_score | sync | Single-page SEO audit with 0–100 score: title, description, canonical, headings, OG, schema, E-E-A-T signals, content depth. Required: url. |
| get_scan_status | sync | Poll the status of an async scan. Required: tool (link_checker | seo_analyzer | sitemap_validator), scan_id. |
Async Tools & Polling
link_checker, seo_analyzer, and sitemap_validator crawl multiple pages and can take 10–120 seconds. They return a scan_id immediately — your agent should poll get_scan_status every 3–5 seconds until status is "done" or "complete".
Typical polling flow:
- Call
link_checker(url: "https://example.com")→ receivescan_id: "abc123" - Call
get_scan_status(tool: "link_checker", scan_id: "abc123")every few seconds - When status is
"done", the result includes full stats: broken, blocked, slow, redirect counts
Example JSON-RPC Calls
All requests are POST https://www.bytewavenetwork.com/api/mcp with Content-Type: application/json.
API Reference
The full OpenAPI spec (REST endpoints, WebSocket frames, schema definitions) is available at:
The MCP discovery endpoint (GET /api/mcp) returns the server name, protocol version, endpoint URL, and a summary of all available tools — useful for agents that support server auto-discovery.