MCP server for AI agents

Install

• Claude Code

  Copy

  claude mcp add --transport http pingtest https://pingtest.si/mcp

• Cursor (~/.cursor/mcp.json)

  Copy

  {
    "mcpServers": {
      "pingtest": {
        "url": "https://pingtest.si/mcp"
      }
    }
  }

• VS Code (.vscode/mcp.json)

  Copy

  {
    "servers": {
      "pingtest": {
        "type": "http",
        "url": "https://pingtest.si/mcp"
      }
    }
  }

• Windsurf (mcp_config.json)

  Copy

  {
    "mcpServers": {
      "pingtest": {
        "serverUrl": "https://pingtest.si/mcp"
      }
    }
  }

• Claude and other apps with custom connectors

  Add a custom connector and paste https://pingtest.si/mcp as the server URL. No sign-in is needed.

Tools

• get_ai_provider_status

  Current state, time to first token (p50/p95, last hour), uptime over 24 hours, 7 days and 30 days, the last probe and open official incidents for each provider.

  Ask: “Is Anthropic or OpenAI having problems right now? My calls keep returning 529.”

• recommend_fastest_ai_provider

  Ranks the providers that are healthy right now by median time to first token, with the reason each other provider was skipped.

  Ask: “Which AI API is fastest right now? Leave out OpenAI.”

• get_ai_provider_history

  Daily uptime, probe counts and average time to first token for one provider over 1 to 90 days, plus official incidents.

  Ask: “Was Gemini down at any point in the last week?”

• probe_network_target

  TCP connect time, ICMP echo or DNS lookup speed for a public host, measured from our edge network. Private addresses are refused.

  Ask: “Can you reach api.example.com on port 443, and how fast does example.com resolve?”

Try it

These buttons send the same JSON-RPC request an MCP client sends to https://pingtest.si/mcp and show the raw response.

Enterprise & Pro monitoring

The free tier allows 20 tool calls per hour. Pro, Scale and Enterprise API keys add monthly allowances for agents and failover routing, plus alerts on your own endpoints.

Questions