How the upmonitor:// Protocol Turns AI Agents into Self-Healing Infrastructure

Updated April 2026: The handoff is now packaged as the Active Incident panel - every degraded monitor surfaces a single copy-pasteable AI prompt at the top of its details page. The protocol scheme described below is one of the two handoff buttons. See the agentic setup guide for the per-tool install (Claude Code, Cursor, Claude Desktop, ChatGPT).

The 3am incident is a rite of passage for every infrastructure engineer. An SSL certificate expires, a DNS record is accidentally deleted, or a botched deployment breaks your security headers.

Standard monitors send an email that sits in an inbox for 40 minutes while your high-traffic application loses thousands of dollars. UpMonitor bridges the gap between cloud alerts and your local AI-native workflow - so triage happens in seconds, not minutes.

The Bridge: Cloud Monitoring to Local Agent

UpMonitor ships two tightly-integrated npm packages that work together to turn alerts into context your AI agents can act on:

@upmonitor/cli - a global CLI for running diagnostics from your terminal or CI/CD pipeline.
@upmonitor/mcp-server - a Model Context Protocol server that lets Claude Desktop, Claude Code, Cursor, and any MCP-compatible agent call UpMonitor's diagnostic suite through natural language.

Together, they let you go from "something's wrong" to "here's what's wrong and where" in one step.

What is the `upmonitor://` Protocol?

The CLI registers a custom URL scheme on your machine - upmonitor:// - just like mailto:// opens your email client. When an alert lands in your Slack or agent UI, clicking the link deep-links straight into your terminal with the URL pre-filled:

upmonitor://check?url=https://example.com

This eliminates the copy-paste dance between dashboard and terminal. Register it once:

upmonitor register

Windows writes to the Registry automatically; macOS and Linux print platform-specific setup steps.

Setting Up Your Diagnostic Workflow

1. Install the CLI

npm install -g @upmonitor/cli

2. Log in

Device-code flow - no API keys to paste:

upmonitor login

The CLI prints a verification code, opens your browser, and polls until you approve. Your token is stored at ~/.config/upmonitor/credentials.json with 0600 permissions.

3. Run a diagnostic

upmonitor check https://example.com

Or pipe the JSON output into anything:

upmonitor check https://example.com --format json | jq '.results.ssl'

For CI/CD:

- name: Post-Deployment Audit
  run: npx @upmonitor/cli check https://$HOSTNAME --ci --fail-on failure

Letting AI Agents Do the Diagnosis

Wire the MCP server into your AI agent and ask questions directly - no context-switching to a browser.

Claude Desktop, Claude Code, Cursor

{
  "mcpServers": {
    "upmonitor": {
      "command": "npx",
      "args": ["-y", "@upmonitor/mcp-server"],
      "env": {
        "UPMONITOR_API_KEY": "up_..."
      }
    }
  }
}

Then ask your agent:

"Is example.com's SSL valid?" → agent calls run_audit with the ssl checker
"What security headers is this site missing?" → agent calls run_audit with securityHeaders
"Check if mysite.com's DNS is healthy" → full DNS + DNSSEC audit

The agent decides which checkers to run, summarizes the output, and suggests fixes - all without you leaving your IDE.

What's Next?

We're building towards a world where "Downtime" is a solved problem. Continuous monitor management via the CLI (monitor add, monitor list) and dedicated MCP tools for status queries (get_monitor_status, add_monitor) are landing in the next release.

Explore the CLI Reference → Learn about MCP Integration →

See it handle a live failure

When a check fails, the Active Incident workflow hands the diagnosis straight to your AI agent for repair.