Skip to content
modsignal

Monitors

A monitor is one watched URL, one plain-language prompt, one check tier, and one schedule. Everything else — snapshots, alerts, evidence — hangs off it.

Creating a monitor

A monitor has six parts, set at creation and editable any time:

  • Name— how it appears in lists and alerts (“Acme pricing”).
  • URL — the public page to watch. Pages behind logins, paywalls, or on private networks are refused at creation; modsignal monitors the public web only.
  • Prompt — the sentence describing what counts as a change. On the semantic and browser-agent tiers this is what the model judges against; on exact diff it documents intent for your teammates.
  • Check tier diff, ai, or agent. See check tiers.
  • Check interval — how often to check, from every 10 minutes (fastest, plan-dependent) to daily. Leave it unset to use your plan's default cadence.
  • Tags— up to 10 free-form labels for organizing a large monitor list (“competitor”, “vendor-risk”).

The first run captures a baseline snapshot and never alerts — there is nothing to compare against yet. Every subsequent check compares the fresh page against the last snapshot. You can trigger a check at any time with Run now, on the monitor page or via the API.

Writing prompts that stay quiet

The prompt is the noise filter. A vague prompt (“watch this page”) makes the semantic tier guess at your intent; a precise one tells it exactly what to ignore. Good prompts name three things — the change that matters, the details that define it, and the noise to skip:

  • Watch this pricing page. Alert me if a price, plan limit, or trial length changes, or a plan is added or removed. Ignore design changes and copy edits.
  • Alert me if any subprocessor is added or removed, or if a processing location moves outside the EU/EEA.
  • Tell me when this changelog announces a breaking change or deprecation for the v2 API. Ignore minor releases.

Prompts are editable without losing history — refining the “ignore” clause after a false alarm is the normal workflow, not a failure.

Templates

The Templates page in the dashboard ships ready-made monitors for common jobs — competitor pricing pages, subprocessor lists, API deprecation notices, marketplace reviews — each with a working prompt and a sensible tier already chosen. Start from one and edit the URL; the prompts encode what we know about each page type's noise profile.

Scheduling and intervals

Checks run on the interval you set, clamped to your plan's fastest allowed cadence: daily on Free, every 15 minutes on Pro, every 5 minutes on Scale, every 2 minutes on Business. The interval picker only offers what your plan allows; if a plan change leaves a monitor on a faster interval than the new plan supports, it simply runs at the plan's fastest instead. Checks are never billed per-run; the interval is a plan feature, not a meter.

Match the interval to the page's rhythm, not to anxiety: legal and subprocessor pages change a few times a year (daily is plenty, and your objection windows are measured in weeks); pricing pages move during launches (hourly during launch season, daily otherwise).

Pausing, resuming, and failure handling

Pause a monitor and it keeps its history, prompt, and schedule but stops checking; resume is one click. Deleting a monitor deletes its snapshots and alert history with it (owners and admins only).

If a monitor fails 5 consecutive checks — the page is gone, the server blocks us, the URL changed — modsignal pauses it automatically and notifies your team on every configured channel, instead of burning checks against a dead page forever. The notice includes the last error; review and resume from the monitor page.