What to watch, and how.

1 · Create your inbound heartbeat

2 · Then curl it from your job

# success — at the end of the job
pg_dump … | gzip | aws s3 cp - s3://… \
  && curl -fsS "$HOLTER_URL"

# failure — alert right away
curl -fsS "$HOLTER_URL/fail"

Miss the window and Holter opens an incident and alerts your channels. Same pattern for queue workers (ping every few minutes) and crons (append the curl to the line).

Prefer email? Switch the heartbeat to a unique address (…@in.holter.sh) and have your app send to it on a schedule. If your mail stops going out, Holter tells you — it watches the whole send pipeline, not just your code.

Create an outbound check

What it catches

• ● Down, or the wrong status code
• ● Up but slow past your latency budget (degraded)
• ● Recovered — with the outage duration

Switch What to check to SSL, Domain, DNS or Redirect — same screen, just a URL, no extra fields. Cert and domain checks warn you days before expiry.

Tip: inside Holter, “What to watch” sets any of these up for you in a click.

Add a dependency

What it catches

• ● A vendor — or one of its components — reports degraded service
• ● A vendor reports a full outage — alerted to your channels
• ● Recovered — the upstream incident cleared

Holter reads each vendor's own published status page every few minutes — it doesn't probe their systems. Works for any page with a public status feed (most SaaS), with built-in support for AWS, Google Cloud and Azure.

Optionally list chosen dependencies on your own status page as “Upstream services.”

Grab the snippet

Live example

• ✓ Self-updating — reflects your real status within a minute
• ✓ Public and self-contained — renders anywhere an image does

1 · Add the MCP server

{
  "mcpServers": {
    "holter": {
      "type": "http",
      "url": "https://app.holter.sh/api/mcp",
      "headers": { "Authorization": "Bearer ${HOLTER_TOKEN}" }
    }
  }
}

Grab a team-scoped token from Settings → MCP and set it as HOLTER_TOKEN. The token is the only thing the agent needs — it is scoped to your Holter team and carries no access to your code, servers, or cloud.

2 · Then just ask

"Set up Holter monitoring for this repo —
read holter.sh/use and follow it."

• ✓ It reads your codebase locally — Holter never gets your source
• ✓ It proposes the monitors and heartbeat code; you review and apply every change
• ✓ It works through the token, never your keys — it can't touch your systems