Catch drift when everything still looks green
A successful run and valid credentials do not guarantee useful data. Fields can disappear, values can become empty, and output can change without triggering an obvious error. Your pipeline may still look healthy while the data it depends on is no longer usable.
$0.02 per check • No subscription
Your pipeline degrades silently.
The run succeeds. The credentials are valid. Yet an expected field has disappeared or the output is unexpectedly empty. Nothing needs to crash for the data your pipeline consumes to stop meeting your expectations.
Catch output schema changes
Check sampled output for missing or renamed fields, even when the run completed successfully.
Check credential validity
See when configured access stops working, alongside your output-quality checks.
Track available quota
Compare reported usage with your configured warning thresholds.
Investigate changes sooner
Use timestamped results and failure details to focus your investigation.
Understand downstream impact
Add your own impact notes so the team knows why a detected change matters.
Review health over time
Use dataset records to follow the health of your configured checks.
The steady pulse of reliable infrastructure.
Datasource Pulse runs on a schedule you choose. It checks configured expectations, compares results with the previous state, and records detected changes with supporting details.
1. Scheduled Trigger
Runs every 1-6 hours, or triggered on-demand by an AI agent.
2. Probe Data
Checks available run history and sampled output, or configured credential and quota signals.
3. State Check
Compares result to previous memory. Detects degradation or recovery.
4. Notify and Record
Sends configured webhook or email notifications for detected state changes. Writes a dataset record.
Context for investigation.
Notifications highlight detected degradation, escalation and recovery. Previous state helps reduce repeated notifications while a detected condition remains unchanged. Dataset records retain the results of each check.
- Includes the specific failure classification (e.g.,
schema_drift) to help focus your investigation. - Echoes your custom impact note verbatim so the on-call engineer knows what broke downstream.
{
"transition": "healthy -> failing",
"connector_id": "output-monitor",
"failure_class": "schema_drift",
"message": "Output schema changed.",
"impact": "Expected fields may be missing downstream",
"metadata": {
"missing_fields": ["record_id"]
}
}Output drift and silent failures
Datasource Pulse checks available Actor run history and sampled output against configured expectations to detect schema drift, unexpected empty output and changes in output volume.
A green run status tells you a run completed, not that its output still meets your expectations.
{"check_type": "actor","connector_label": "Output monitor","state": "failing","failure_class": "schema_drift","message": "Output schema changed.","metadata": {"missing_fields": ["record_id"],"new_fields": ["id"]}}
Expected fields were missing from the sampled output.
Credential & Quota Health
Additional checks cover request failures, credential validity, available quota and optional run-cost changes. Results include timestamps and supporting details so you can see what was observed.
Check Credential Validity
Check whether configured credentials are accepted. Valid access is one signal, not proof that the data is useful.
Custom Quota Soft Caps
Compare reported usage against your configured budget line, where the service exposes quota information.
Send your monitoring requirements for a coverage and setup review.
Monitor your own HTTP endpoint
Use a generic_http check to compare an HTTP response with your configured success condition. Configure your URL and optional authentication in the Input tab.
"vendor": "generic_http",
"url": "https://your-service.example/health",
"successStatusCode": 200
Agent-first architecture.
AI agents build autonomous data pipelines. Datasource Pulse gives them structured results for detected output drift and other configured health checks.
API & MCP Ready
Discoverable and callable as a tool via the Apify API and MCP. Takes a structured JSON input manifest, returns exactly one status record per connector.
Agent access docs →Machine-Readable Errors
No parsing human error strings. A stable failure_class enum (dead_token, schema_drift, rate_limited) lets agents branch logic automatically.
x402 Payments · Live
Agents buy health checks from our API today: $0.02 in USDC on Base, no account of any kind. Listed in Coinbase’s x402 Bazaar, settled on-chain, verifiable by anyone.
Radically simple pricing.
Pay for the check, not the result. Each attempted check is billed, whether its result is healthy, warning or failing.
Pay per event
No subscriptions. No per-run minimums.
- Charge applies on attempt (failures count)
- Subset recheck: check just one configured source before an expensive job
- Set a hard max cost ceiling per run
- x402 agentic payments supported (USDC on Base)
Frequently asked questions
How do I get an alert when a data feed goes stale?
Watch freshness, not uptime: a stale feed still answers 200. The manual route is a daily job comparing the feed’s newest timestamp against a threshold set just past its update cadence. The Actor checks review available run history for reported zero-yield streaks, optional volume drops against a learned baseline, and Actor staleness. HTTP checks test your configured response condition; feed-timestamp comparisons remain client-side. Checks cost $0.02 each. Full walkthrough of both routes: how to get an alert when a data feed goes stale.