API Breaking-Change Impact Radar

Monitor any public API for breaking changes, deprecations and sunsets — before they break your production. The actor diffs OpenAPI specs with oasdiff (the reference engine for spec-level breaking-change detection) and classifies GitHub releases / RSS changelog entries into audit-ready impact events.

Use this tool when...

• You integrate third-party APIs (Stripe, GitHub, Twilio, OpenAI, Discord, Slack…) and need to know the day a breaking change or deprecation appears — not the day your clients call you.
• You run a platform or agency and must route each change to the right owner with evidence and a recommended action.
• An AI agent needs a structured, MCP/API-consumable feed of API impact events (each event carries evidence_url, severity, recommended_action).

How it works

1. openapi sources (most precise): the spec is fetched and hashed. On change, the previous and new specs are diffed with oasdiff breaking --format json. Every entry is normalized: path, operation, severity (critical = breaking, warning = deprecation, info = additive).
2. github-releases sources: new releases since the last run are classified by rule tiers (breaking / deprecation / security / additive keywords).
3. rss sources: new feed entries classified the same way.

State lives in the run's key-value store, so the first run records a baseline and detection starts on the next change. Schedule the actor (daily/weekly) for continuous radar coverage.

Input

Field	Type	Description
apis	array	Required. APIs to monitor: {name, type, url, owner?}. type = openapi (raw spec URL), github-releases (owner/repo or repo URL), rss (feed URL).
webhookUrl	string	Optional. POST endpoint receiving a JSON summary at the end of each run (Slack/Discord/Teams compatible text field included).
dryRun	boolean	Default true. When true, no pay-per-event fee is charged — full detection runs, billing is simulated. Set to false in production.
maxApis	integer	Safety cap per run (default 25).

Example input

{
  "apis": [
    {"name": "github", "type": "openapi", "url": "https://raw.githubusercontent.com/github/rest-api-description/main/descriptions/api.github.com/api.github.com.json", "owner": "platform-team"},
    {"name": "stripe", "type": "openapi", "url": "https://raw.githubusercontent.com/stripe/openapi/master/openapi/spec3.json", "owner": "payments"},
    {"name": "my-provider", "type": "github-releases", "url": "owner/repo"}
  ],
  "webhookUrl": "https://hooks.slack.com/services/XXX/YYY/ZZZ",
  "dryRun": false
}

Output

One dataset item per detected change:

Field	Description
api / owner	API name and the responsible person/team you declared
detected_at	ISO-8601 detection timestamp
change_class	breaking · deprecation · security · additive · baseline · monitor-error
severity	critical · warning · info
title / summary	What changed, human-readable
path / method	Affected endpoint (openapi sources)
evidence_url	Public proof of the change — auditable
detector	Engine that caught it (oasdiff, rule:<tier>, baseline)
recommended_action	Concrete next step

A run-level SUMMARY record (counts by class + top critical changes) is stored in the output key-value store.

Pricing (pay-per-event, transparent)

Event	What you pay for
run-started	Flat per run — covers fixed compute so you never pay per tiny check
api-checked	Per API monitored per run
breaking-change-detected	Per breaking/deprecation event found — the unit of value
critical-alert	Per critical-severity event with evidence

Compute is a few seconds of CPU per API (no browser, no proxy), so runs cost cents. Set dryRun: true to test your configuration for free.

Scheduling

Use Apify Schedules (daily at 06:00 recommended) so baselines stay fresh and every provider change is caught within 24h.

Limits & honesty

• openapi detection is exact only when the provider maintains a public spec. Without a spec, classification is keyword-based and can miss or over-flag entries — treat warning items as review candidates.
• GitHub release fetching uses the unauthenticated public API (60 req/h shared). For heavy multi-repo setups, run less frequently.
• The actor never guarantees exhaustive coverage: it is a radar, not an audit. Verify critical items against the provider's official changelog before acting.

AI-agent ready · API + MCP ready · Built on oasdiff.