The MCP server authenticates every request using your Webtools account email and an API key. You need both before connecting any agent.

1. Sign in to Webtools and go to your Profile page.
2. Scroll to the API Key section and click Generate API Key.
3. Copy the key — it is shown only once. If you lose it, generate a new one (the old key is immediately invalidated).

Kiro reads MCP server configuration from a JSON file in your workspace or user home directory. Open (or create) the file at one of these paths:

• Workspace-level (applies to one project): .kiro/settings/mcp.json
• User-level (applies to all projects): ~/.kiro/settings/mcp.json

Add the following block inside the mcpServers object, replacing the placeholder values with your real credentials:

{
  "mcpServers": {
    "webtools": {
      "url": "https://webtools.ayalr.com/mcp",
      "headers": {
        "X-Email":   "[email protected]",
        "X-API-Key": "your_api_key_here"
      }
    }
  }
}

If you already have other MCP servers configured, merge the webtools block into your existing mcpServers object — don't replace the whole file:

{
  "mcpServers": {
    "some-other-server": { "..." },
    "webtools": {
      "url": "https://webtools.ayalr.com/mcp",
      "headers": {
        "X-Email":   "[email protected]",
        "X-API-Key": "your_api_key_here"
      }
    }
  }
}

After saving the file, Kiro reconnects automatically. You can also trigger a reconnect from the MCP Server view in the Kiro feature panel, or by searching MCP in the command palette.

Open the Claude Desktop configuration file for your operating system:

Add the same webtools block as shown in Step 2. If the file doesn't exist yet, create it with this content:

{
  "mcpServers": {
    "webtools": {
      "url": "https://webtools.ayalr.com/mcp",
      "headers": {
        "X-Email":   "[email protected]",
        "X-API-Key": "your_api_key_here"
      }
    }
  }
}

Restart Claude Desktop after saving. The Webtools tools will appear in the tool picker (the icon) in any new conversation.

Once your agent is configured, confirm it can reach the server before running any tests.

Option A — Ask the agent directly:

The agent will call get_quota and return your plan name, total monthly allowance, calls used, and remaining calls. This tool never consumes quota, so it's safe to call as many times as you like.

Option B — Test the endpoint directly with curl:

curl -X POST https://webtools.ayalr.com/mcp \
  -H "Content-Type: application/json" \
  -H "X-Email: [email protected]" \
  -H "X-API-Key: your_api_key_here" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "get_quota",
      "arguments": {}
    }
  }'

You should receive a JSON-RPC response like:

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [{
      "type": "text",
      "text": "{\n  \"plan_name\": \"Pro\",\n  \"total_calls\": 1000,\n  \"used_calls\": 42,\n  \"remaining_calls\": 958\n}"
    }]
  }
}

You can also GET https://webtools.ayalr.com/mcp (no auth required) to see the server's discovery metadata.

With the server connected, try these prompts to get started. The agent will automatically discover and call the right tools.

The MCP server exposes 9 tools. The agent selects and calls them automatically based on your prompt — you don't need to name them explicitly.

Quota note: Each tool call that reads or writes data counts as one API call against your monthly allowance. Use get_quota to check your remaining balance before running bulk operations. Upgrade your plan on the pricing page to increase your allowance.

These prompts work with any MCP-compatible agent. The agent figures out which tools to call and in what order.

Performance monitoring

• "Check my quota, then run a performance test on site ID 1 and tell me the results when they're ready."
• "Compare this week's Lighthouse scores to last week's for all my sites. Highlight any URLs where performance dropped by more than 10 points."
• "Which of my pages has the worst Largest Contentful Paint over the last 30 days?"
• "After my next deployment, run a lab test on site 2 and fail if LCP exceeds 2.5 seconds."

SEO & crawl audits

• "Run a crawl on crawler ID 8 and give me a prioritised list of issues to fix."
• "How many broken links does my site have? Which pages are they on?"
• "Check the latest crawl results and tell me if there are any pages with duplicate H1 tags or missing titles."
• "Compare the last two crawls and tell me if the number of non-200 responses has increased."

Uptime & monitoring

• "Are all my uptime monitors healthy right now? Show me any that failed in the last 24 hours."
• "What's the average response time for monitor ID 12 over the last week?"
• "Check my server metrics for the last hour and tell me if CPU or memory looks abnormal."

Quota-aware automation

• "Check my quota first. If I have more than 20 calls remaining, run tests on all my sites and summarise the results."
• "How many API calls do I have left this month? Is it enough to run a full audit of all my sites and crawlers?"

The Webtools MCP server implements the Model Context Protocol over Streamable HTTP. When your AI agent needs to monitor a website, it sends a JSON-RPC 2.0 request to POST https://webtools.ayalr.com/mcp. The server authenticates the request, executes the tool, and returns the result — all within a single HTTP round-trip.

Technical details

Autonomous Website Monitoring Using AI Agents and MCP

The Webtools MCP server bridges the gap between AI language models and real website monitoring data. By implementing the Model Context Protocol over a simple HTTPS endpoint, it lets any MCP-compatible agent — Claude, Kiro, GPT-4 with tool use, or your own custom agent — query live performance metrics, trigger Lighthouse audits, run SEO crawls, and check uptime status without any manual dashboard interaction.

What is MCP and why does it matter for monitoring?

MCP (Model Context Protocol) is an open standard that lets AI agents discover and call external tools at runtime. Instead of hard-coding API calls into your automation scripts, you describe your tools once and any MCP-compatible agent can use them. For website monitoring, this means you can ask an agent in plain English to investigate a performance regression, triage broken links from a crawl, or summarise uptime failures — and the agent will figure out which tools to call, in what order, and how to interpret the results.

Core Web Vitals monitoring via AI

The run_site_test and get_site_results tools give agents direct access to Lighthouse-based lab test data: LCP, CLS, TBT, FCP, Speed Index, and TTFB for every URL on your site. An agent can compare scores across deployments, flag regressions against your own thresholds, and produce a plain-language summary — all without you opening a single dashboard.

Automated SEO crawl analysis

The run_crawl and get_crawl_details tools let agents trigger and analyse full-site crawls. The raw crawl data includes every visited URL, its HTTP status code, H1 tags, canonical tags, and a summary of issues: broken links, 4xx/5xx errors, missing titles, duplicate canonicals, and mixed content. An agent can prioritise these issues by severity, cross-reference them with performance data, and produce an actionable SEO audit report.

Uptime and RUM data in natural language

The list_monitors and get_monitor_results tools expose uptime check history, Real User Monitoring session data, and server metrics. An agent can detect anomalies, correlate uptime failures with deployment events, and summarise SLA compliance — all from a single prompt.