API Endpoint Health & JSON Schema Drift Monitor

Check public HTTP/JSON APIs for downtime, latency regressions, missing required fields, and unexpected response-shape changes. Designed for developers, automation teams, data pipeline operators, and AI agents that rely on third-party APIs. This Actor runs one-off or on an Apify schedule and returns structured machine-readable health events.

What this API monitor checks

Each run checks up to 20 public HTTP/HTTPS endpoints, measuring HTTP status and response time, validating JSON where requested, verifying configured JSON paths such as data.items[0].id, and fingerprinting a bounded JSON structure. Results include the HTTP status, latency, required-path failures, JSON validity, schema fingerprint, change type, and a timestamp. An endpoint is unhealthy if its expected status, JSON requirement, required fields, or latency threshold is not met.

The monitor stores a named baseline in its own Apify key-value store. A first check emits INITIAL. Later checks distinguish RECOVERED, DEGRADED, STATUS_CHANGED, SCHEMA_CHANGED, CONTRACT_CHANGED, and UNCHANGED. This is useful for detecting when a vendor silently removes an API field, when a service becomes slow, or when a workflow's upstream data source fails.

Who uses it, and why checks recur

• AI-agent operators: verify public tools and JSON data sources before workflows depend on them.
• Developers: detect API status and shallow response-schema drift after vendor releases.
• Data engineering teams: monitor the external JSON paths consumed by ETL jobs.
• Agencies and SaaS teams: schedule recurring checks across customer integrations.
• Compliance/operations teams: keep an auditable dataset of periodic API checks.

These checks are valuable because APIs can change between deployments without warning. Schedule hourly or daily based on the impact of a failure. Apify schedules run the Actor; it is not a continuous daemon between scheduled runs.

Example input

{
  "endpoints": [
    {
      "name": "Apify Store API",
      "url": "https://api.apify.com/v2/store?limit=1",
      "expectedStatus": 200,
      "requiredJsonPaths": ["data.items"]
    }
  ],
  "monitorKey": "my-public-api-watchlist",
  "checkJson": true,
  "emitOnlyChanges": false,
  "latencyWarnMs": 3000,
  "timeoutMs": 8000
}

emitOnlyChanges=false (default) writes a dataset row for every endpoint check, including healthy unchanged checks, so a monitoring dashboard can show the latest status. Set it to true to emit only initial or changed events while still updating the baseline. Only emitted dataset rows trigger the proposed per-result charge; Apify may also apply a minimal run-start charge.

Output and integrations

Use the default dataset for JSON/CSV export or consume results through the Apify API, webhooks, and supported automation integrations. The Actor also writes a SUMMARY record with counts of healthy, unhealthy, changed, and emitted checks. Output rows contain name, url, httpStatus, expectedStatus, latencyMs, healthy, jsonValid, missingPaths, schemaFingerprint, changeType, changed, error, and checkedAt.

Safety, limits, and interpretation

Only public, unauthenticated HTTP/HTTPS endpoints are supported. Local, private, and reserved network addresses are blocked, as are credentials embedded in URLs and common API-key query parameters. Requests use GET and follow at most two redirects; each destination is checked. Responses are capped at 500 KB and time out after at most 15 seconds. The Actor does not send credentials or write to monitored endpoints.

A schema fingerprint compares a bounded sample of the response shape, not a complete OpenAPI or JSON Schema contract. Dynamic array shapes can cause expected changes, and a healthy result does not prove the endpoint is secure or correct. DNS rebinding defenses are best-effort; do not treat this as a network security boundary. This is operational monitoring, not penetration testing or an SLA guarantee.

Pricing and cost control

The intended listing price is $0.002 per emitted endpoint-check dataset item ($2 per 1,000). A run with 10 endpoints emits up to 10 billable rows when emitOnlyChanges=false. Use emitOnlyChanges=true to pay only for initial/changed results and a small endpoint list for initial testing. No paid upstream Actors, residential proxies, or API subscriptions are required by the code. Platform compute and customer run limits still apply.