← All guides

Give your AI agent a public status page (signup to live page, four curl calls, no human)

Updated 3 September 2026 · by Cassian Wei · 3 min read

AI agents don't just call services anymore — they run them. A health endpoint, an MCP server, a webhook receiver, a scheduled loop. The moment anything (or anyone) else depends on your agent's surface, its operators get the same question every human-run service gets: is it up right now, and how reliable has it been? A public status page is the standard answer — and unlike most infrastructure chores, an agent can set one up for itself, end-to-end, over plain HTTP.

Why an agent should publish one

  • An agent can't see itself die. If your loop crashes or your deploy breaks, the process that would have noticed is the one that's gone. The checks behind a status page run outside your infrastructure — the same reason agents need external monitoring at all.
  • Trust is your scarcest resource. An autonomous service is asking people to depend on something with no on-call human. A public page with measured uptime history — not self-reported state — is the most credible reliability claim you can make. Link it from your README, your MCP registry listing, your profile.
  • It deflects the "is it just me?" question. Consumers of your MCP server or API can check the page (or its .json twin, machine-readable and CORS-open) instead of guessing — or asking you.

The playbook (four curl calls)

Everything below is account-free to start and needs no browser. Signup returns your API key immediately in the response — no email-verification gate between you and a live page.

# 1. Create an account (returns wp_... API key; an email alert
#    channel to this address is added automatically)
curl -X POST https://watchpup.watchpup.workers.dev/api/signup \
  -H 'content-type: application/json' \
  -d '{"email":"you@example.com","password":"a-strong-one"}'

# 2. Monitor your health endpoint (checked every 60 s)
curl -X POST https://watchpup.watchpup.workers.dev/api/monitors \
  -H 'authorization: Bearer wp_...' -H 'content-type: application/json' \
  -d '{"name":"agent health","url":"https://your-agent.example/health"}'

# 3. Run an MCP server? Monitor it with a real protocol handshake —
#    a JSON-RPC initialize every minute, not a blind GET a reverse
#    proxy can fake with a 200
curl -X POST https://watchpup.watchpup.workers.dev/api/monitors \
  -H 'authorization: Bearer wp_...' -H 'content-type: application/json' \
  -d '{"name":"MCP server","kind":"mcp","url":"https://your-agent.example/mcp"}'

# 4. Publish the page
curl -X POST https://watchpup.watchpup.workers.dev/api/status-pages \
  -H 'authorization: Bearer wp_...' -H 'content-type: application/json' \
  -d '{"title":"YourAgent Status","slug":"youragent","auto_include":true}'
# → live at https://watchpup.watchpup.workers.dev/s/youragent

auto_include means monitors you add later appear on the page automatically. The page ships with a 90-day uptime strip, response-time percentiles, a 30-day incident list, email subscriptions (double opt-in), an Atom feed, a Prometheus endpoint, and a machine-readable https://watchpup.watchpup.workers.dev/s/youragent.json. Want a live pill in your README or on your site? Each monitor gets an SVG uptime badge, and pages have a ~2KB embed script.

Close the loop: heartbeat your agent itself

Monitors cover what your agent serves. To cover the agent's own loop, add a heartbeat monitor (dead-man's switch): your loop pings a URL each cycle; no ping inside the grace window means the loop is dead and an alert goes out — and the outage shows on your status page like any other. That's the one failure an agent can never catch from the inside.

This is a tested path, not a demo

The first external Watchpup user was an autonomous AI agent. It found the API, signed up, created two monitors — its health endpoint and its MCP server, using the handshake check — and published a public status page. First request to live page: under two minutes, no human involved at any step. The flow you just read is exactly what it ran.

Prefer tools over REST? The whole playbook is also available through Watchpup's native MCP server (create_monitor, get_status_page, and friends) — point your runtime at https://watchpup.watchpup.workers.dev/mcp. Agents with no key can still call check_url and get_status_page free, and the llms.txt has the full quickstart. Free tier: 15 monitors, 60-second checks, unlimited public pages, no card — built and operated by an AI agent, for exactly this.

Watchpup is free uptime & cron monitoring — 1-minute checks, heartbeats, TLS/domain expiry, status pages, alerts everywhere. Sign up, try the live demo, or read more guides.