MCP integration

Connect AI assistants to PingChange via Model Context Protocol—endpoint, authentication, and what the tools can do.

Model Context Protocol (MCP) lets compatible AI assistants and automation clients query your PingChange workspace safely: list monitors, read recent changes, manage channels, or kick off checks — with the same permissions implied by your workspace API key.

Prerequisites

  1. Workspace API keyDashboard → Settings → API keys; copy or regenerate there. Every MCP request uses this key as a Bearer token. Details: Workspace API key.
  2. MCP pageDashboard → MCP shows your endpoint for the environment you are logged into, plus copy-ready JSON for common clients.

The in-app page builds the same URLs and config blocks described below, with your real key filled in. Prefer Copy on that page over retyping secrets.

MCP endpoint (streamable HTTP)

PingChange exposes MCP over streamable HTTP at:

https://pingchange.com/api/llm/mcp

The path is always /api/llm/mcp. When you are signed in, Dashboard → MCP shows the URL for your current origin.

Authentication

Every request must send your workspace API key as a Bearer token:

Authorization: Bearer YOUR_WORKSPACE_API_KEY

Paste the key from Settings → API keys; do not commit it to public repositories or share config files that contain the live secret.

If your client supports streamable HTTP, point it at the MCP URL above and attach the Bearer header.

Many tools expect a top-level mcpServers object. The outer key name is arbitrary (rename "mcpServers" if your client uses a different shape). The inner server id ping-change is only a label—you can rename it to match your config file.

{
  "mcpServers": {
    "ping-change": {
      "url": "https://pingchange.com/api/llm/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_WORKSPACE_API_KEY"
      }
    }
  }
}

Replace the URL with your origin in non-production environments. Replace YOUR_WORKSPACE_API_KEY with the key from Dashboard → Settings → API keys.

Cursor / Claude Desktop: open your MCP config file, paste the block (or merge the "ping-change" entry into existing mcpServers), save, and restart the client if it does not reconnect automatically.

Tool names and JSON shapes can vary by client; follow the assistant’s guide for adding an HTTP MCP server when field names differ slightly.

stdio via mcp-remote

Some assistants only support stdio MCP servers. Use mcp-remote to proxy the same streamable HTTP endpoint. The config runs npx with your URL and passes auth via --header (requires an mcp-remote build that supports --header):

{
  "mcpServers": {
    "ping-change": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://pingchange.com/api/llm/mcp",
        "--header",
        "Authorization: Bearer YOUR_WORKSPACE_API_KEY"
      ]
    }
  }
}

Again, swap the URL and Bearer value for your environment and key. Dashboard → MCP provides a one-click Copy JSON for this block with your credentials already inserted.

What assistants can do (overview)

Once connected, tools typically allow your assistant to:

  • Summarize the workspace — high-level counts of monitors, channels, and activity.
  • List monitors — names, URLs, types, schedules, and status (including paused monitors).
  • Read recent changes — the same kind of information as CSV/JSON exports (summaries, links, optional screenshots).
  • List notification channels — what destinations exist for alerts.
  • Update a monitor — rename, pause or resume, change schedule or regions, or edit the AI prompt fields the API exposes.
  • Run a check now — queue an immediate run for a monitor.
  • Resend alerts — re-deliver notifications for a past change (recovery after fixing a channel).
  • Link channels to monitors — attach or detach destinations.
  • Create a channel — add a new email, Slack, or other destination (then link it to monitors as needed).

Exact capabilities evolve with the product; use the MCP connection together with the in-app dashboards for critical operations.