Globalping MCP Server for Network Measurements

Overview

The Globalping MCP Server exposes Globalping’s global network measurement platform over the Model Context Protocol (MCP). It lets language models and MCP-capable clients (for example, GPT-based or Claude assistants) request and interpret network tests—ping, traceroute, DNS, MTR, HTTP—using natural-language workflows. By presenting structured measurement capabilities and parameter metadata through MCP, the server enables AI agents to decide which tests to run, where to run them from, and how to compare results.

This is useful for interactive diagnostics, automated performance checks, and augmenting conversational agents with real, live network telemetry from thousands of globally distributed probes. Developers can integrate the remote MCP server into existing MCP tooling without hosting their own measurement backend, and can authenticate either with OAuth (handled by the server) or an API token to receive higher rate limits and account-scoped access.

Repository: https://github.com/jsdelivr/globalping-mcp-server

Features

• Global network measurements from thousands of probes worldwide
• Support for ping, traceroute, DNS, MTR, and HTTP tests
• MCP-compatible interface for LLMs and other MCP-aware clients
• Two authentication modes: OAuth (automatic) and API token (Authorization header)
• Parameter descriptions to help AI agents choose measurements intelligently
• Comparative analysis primitives (compare results across targets or regions)
• Streamable HTTP and legacy SSE transports for compatibility with different clients

Installation / Configuration

The Globalping MCP Server is available as a remote MCP endpoint. You do not need to run a server locally to use it.

Endpoints and transports:

Common configuration examples:

• Add server with an Authorization header (API token):

# Example for Gemini CLI (manual add)
gemini mcp add globalping https://mcp.globalping.dev/mcp --header "Authorization: Bearer YOUR_GLOBALPING_API_TOKEN"

• Node-based bridge (mcp-remote) for clients that require a local stdio MCP server (e.g., Claude Desktop):

{
  "mcpServers": {
    "globalping": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.globalping.dev/sse"]
    }
  }
}

• Cursor-style config (streamable HTTP):

{
  "mcpServers": {
    "globalping": {
      "url": "https://mcp.globalping.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_GLOBALPING_API_TOKEN"
      }
    }
  }
}

Anthropic Console (adding as a custom tool):

• Tool Name: Globalping
• Description: Run network tests from locations worldwide
• Tool URL: https://mcp.globalping.dev/mcp (or /sse for SSE)

Authentication

• OAuth: Handled transparently by the remote server when you use the generic endpoint without an explicit token; the server will prompt to authorize where supported.
• API Token: Generate at https://dash.globalping.io (Tokens section). Provide the token as an HTTP Authorization header (“Authorization: Bearer ”) in your MCP client configuration to use token-based auth.

  Available Resources

  The MCP server advertises structured capabilities and parameters so clients and LLMs can discover:

  • Measurement types: ping, traceroute, DNS, MTR, HTTP
  • Parameter metadata: target (IP/hostname), probe selectors (country, ASN, city), packet count, timeout, TCP/UDP options for traceroute, DNS record type, HTTP method/headers
  • Comparative tools: run the same test from multiple regions or compare two targets side-by-side
  • Live streaming of results via streamable HTTP or SSE transports

  Clients that consume MCP will receive machine-readable descriptions for each tool and parameter so automated agents can build valid measurement requests.

  Use Cases

  • Regional latency troubleshooting: “From probes in Frankfurt, New York, and Singapore, run a 10-ping test to example.com and report median latency and packet loss.”
  • Multi-target comparison: “Compare traceroute paths from US and EU probes between service-a.example and service-b.example and highlight differences in AS hops.”
  • DNS validation: “From DNS resolvers in APAC, query A and AAAA records for mydomain.dev and verify consistency of responses and TTLs.”
  • Incident triage in a chat assistant: Integrate the MCP server into a support assistant so it can run targeted measurements during a support conversation and explain the findings in natural language.
  • Automated monitoring playbook: Use an LLM orchestration agent to run scheduled MTR tests, interpret the trends, and create alerts when latency or packet loss thresholds are crossed.

  Tips for Developers

  • Prefer the streamable HTTP endpoint for modern MCP clients; use the SSE URL where a client expects legacy SSE transport.
  • If integrating into CI or automation, use an API token in headers to avoid interactive OAuth flows and get predictable rate limits.
  • Inspect the MCP capability descriptions programmatically so your LLM prompts can automatically select required parameters (region, probe filters, measurement type) instead of hardcoding them.

  For implementation details and source, see the repository: https://github.com/jsdelivr/globalping-mcp-server

  Tags:ai-mlmcpmcp servermcp-serverremote mcp serverremote-mcp-server