Back to blog
Product

Manage Your Dead Man's Switches from Claude, Cursor, or Any AI Assistant

Matt5 min read

Your AI Assistant Can Manage Your Monitors Now

TL;DR: The DeadPing MCP server exposes 11 tools that let AI assistants like Claude, Cursor, and Windsurf create monitors, check status, search captured output, list anomalies, and investigate incidents through natural language. Setup takes 30 seconds with an npm install and an API key.

We shipped an MCP server for DeadPing. If you use Claude, Cursor, Windsurf, or any other AI tool that supports Model Context Protocol, you can now manage your dead man's switch monitors without leaving your editor or chat window.

No new dashboard tab. No context switching. Just tell your AI assistant what you want and it does it.

What You Can Actually Do

The MCP server exposes 11 tools. Here's what that looks like in practice:

Natural language → MCP tool calls
"Show me which monitors are down right now"
→ list_monitors(status: "down")

"Create a monitor for my nightly ETL job that runs every day at 2am"
→ create_monitor(name: "nightly-etl", cron_expression: "0 2 * * *")

"What happened during last night's incident?"
→ list_incidents(status: "resolved") → get_incident(id: "...")

"Search for 'connection refused' in recent job output"
→ search_output(query: "connection refused")

"Pause the billing-sync monitor while I deploy"
→ update_monitor(id: "...", paused: true)

"Resume it"
→ update_monitor(id: "...", paused: false)

The AI handles the tool selection, parameter mapping, and result formatting. You just ask questions in plain English.

Why This Is Useful (Not Just Cool)

The obvious pitch is convenience. But the real value is in the workflows it unlocks:

Incident Investigation Without Context Switching

You're debugging a production issue in your terminal. Something is broken and you suspect a background job stopped running. Instead of opening DeadPing in a browser, finding the right monitor, checking the timeline, and reading through captured output – you ask your AI:

In your editor
"Is the payment-processor monitor healthy?
If not, show me the last 5 pings with their exit codes and output."

Your AI calls list_monitors, finds the right one, calls list_pings with the monitor ID, and shows you the exit codes, durations, and captured stderr – all without leaving your editor. You stay in flow.

Monitor-as-Code in Your Deploy Scripts

You're adding a new cron job to your infrastructure. With MCP, your AI can create the DeadPing monitor for it in the same session where you write the code:

Conversation with your AI
You: "I just added a new cron job to clean up expired sessions
      every hour. Create a DeadPing monitor for it."

AI: Created monitor "session-cleanup" with schedule "0 * * * *"
    and 5-minute grace period.

    Ping URL: https://deadping.io/api/ping/abc123...

    Add this to the end of your cleanup script:
    curl -fsS "https://deadping.io/api/ping/abc123?exit_code=$?"}

The monitor exists before you even push the code. No separate setup step to forget about.

Bulk Triage During Outages

When something goes wrong and multiple monitors are down, the last thing you want to do is click through a dashboard one monitor at a time. With MCP:

During an outage
"List all monitors that are currently down.
For each one, show me when it last pinged
and what the captured output says."

Your AI fetches all the data in one shot and presents it in a readable format. You can immediately see if the failures are related (same error across all monitors) or independent.

Setup Takes 30 Seconds

Generate an API key from your dashboard settings, then add the MCP server config to your AI tool:

Claude Desktop / Claude Code config
{
  "mcpServers": {
    "deadping": {
      "url": "https://deadping.io/api/mcp",
      "headers": {
        "X-API-Key": "dp_live_your_key_here"
      }
    }
  }
}

That's it. No npm install, no local server to run. DeadPing hosts the MCP server at deadping.io/api/mcp using the Streamable HTTP transport, so it works with any MCP-compatible client out of the box.

See the full MCP docs for setup instructions for Cursor, Windsurf, and other clients.

The 11 Tools

For reference, here's everything the MCP server can do:

ToolWhat it does
list_monitorsList all monitors, optionally filtered by status
get_monitorGet detailed info for a specific monitor
create_monitorCreate a new monitor with name, schedule, and grace period
update_monitorUpdate settings or pause/resume a monitor
delete_monitorDelete a monitor and its ping history
list_pingsView recent pings with exit codes, duration, and output
search_outputFull-text search across captured job output (Pro+)
list_incidentsView open or resolved incidents
get_incidentGet incident details with affected monitors and timeline
update_incidentAdd root cause and action items to an incident
get_accountView current plan, monitor count, and usage

Available on All Plans

MCP server access is included on Free, Pro, and Business plans. The only tool that requires a paid plan is search_output, which needs Pro or Business for the underlying full-text search capability.

If you're already using DeadPing, generate an API key and try asking your AI assistant about your monitors. If you're not, sign up free – 20 monitors, no credit card.

Start monitoring in 60 seconds

Free forever for up to 20 monitors. No credit card required.

Get Started Free