A website change detection API should tell your app what happened
Five REST calls to create a monitor, and a signed webhook that carries evidence rather than markup.
3 min read
A website change detection API watches a public page, a topic or a social account and posts a structured event when the content that matters moves. AyeWatch runs one at https://ayewatch.ai/api/v1, and its job is to spare your application a second fetch: the event should say what happened, not only that something did.
What the API does
Five REST calls cover a monitor's life: create, list, read, update and delete. The name you send decides the type. Pass a URL and you get a web page monitor. Pass plain text and AyeWatch follows that subject across the web. Pass a source object and it follows one public feed, such as an X profile or a subreddit.
Each topic also takes an interval, a description and an active flag. The description is the rule in plain English, such as "tell me when this page adds a region". For a precise condition, triggers state an entity, an attribute, an operator and a value, with operators such as changed, crossed_above and crossed_below.
Put one page in one topic
One broad topic for a whole site is a common first mistake. When it fires, nobody can tell which page moved. Separate topics keep the history readable and let you pause one noisy source without touching the rest.
What a webhook payload should carry
A receiver should be able to route, log and act without opening the page. Six things earn their place:
- Two identities. The monitor it belongs to, and a unique id for this update. AyeWatch sends topic_id and update_id; the second is what you store to drop a repeated delivery.
- A short summary written for a person, not a diff dump. That is the headline field.
- The body of the change, in content, plus extra detail in content_detail.
- A timestamp, created_at, in ISO 8601.
- A confidence and a novelty signal, where the service provides one. AyeWatch includes score and novelty_class.
- Evidence. Quotes with their source URLs let a reviewer confirm the claim in seconds.
Then a signature. Every AyeWatch delivery carries X-AyeWatch-Signature, an HMAC-SHA256 digest of the raw body. Compute the same digest with your secret, compare in constant time, and reject what does not match. Parse only after that check passes.
How teams use it
| Team | Watched source | Reaction to the event |
|---|---|---|
| Pricing | A competitor's plan page | Post the change in a pricing channel, open a review task |
| Support | Help centre and changelog | Draft a release note before tickets arrive |
| Compliance | A regulator's notices page | File the quote and link with the case |
| Comms | A public account or hashtag | Brief the social lead within the hour |
The pattern repeats. A person used to check a page. Now a rule checks it, and a small program puts the result where someone will see it.
Build for a receiver that can fail
AyeWatch retries a non-2xx response with exponential backoff, up to 10 attempts, and abandons a request after 10 seconds. Deliveries can arrive out of order, and the documentation says to expect new fields over time. Acknowledge fast, queue the work, and keep update_id in a unique index.
What the API does not do
It watches public sources. It does not log in to a private dashboard, push past a bot wall, or buy the thing a price page lists. Keep those expectations out of the design and the integration stays small.
Keys come with paid plans, and each key is limited to 120 requests a minute by default, with rate limit headers on every authenticated response. Push and the in-app feed are included on every plan. Email, Slack, Discord, webhooks and the API are paid features, billed in monitoring credits, with paid plans from $9 a month and a 7 day trial on the web.
FAQ
Do I need my own server?
For a webhook, yes: something public over HTTPS that answers within 10 seconds. Push, email, Slack and Discord need no server of yours.
Can one endpoint receive every monitor?
Yes. Each account has one webhook, and every monitor reports to it. Route on topic_id.
What should I do with a failed delivery?
Return a 2xx once you have stored the event, and let your own queue do the rest. A slow handler will time out even though AyeWatch retries.
Start with one page you check by hand, send its events to a request bin, and read the payload before pointing it at production. AyeWatch shows a free preview of how a monitor reads a page, and the API reference sits at ayewatch.ai/documentation.
FAQ
Do I need my own server?
For a webhook, yes: something public over HTTPS that answers within 10 seconds. Push, email, Slack and Discord need no server of yours.
Can one endpoint receive every monitor?
Yes. Each account has one webhook, and every monitor reports to it. Route on topic_id.
What should I do with a failed delivery?
Return a 2xx once you have stored the event, and let your own queue do the rest. A slow handler will time out even though AyeWatch retries.
Originally published on claudegpt.si