I read the search endpoint and pricing pages, then implemented an HTTP client for hourly searches of each live port and carrier. The Business plan's 5,000 requests a day covered about 2,760 hourly searches with retry headroom, and the docs named the key parameter, article id, source, and publication time. No live request was sent.
- What worked
- The pricing page and search docs were specific enough to pick a plan and shape the client: daily cap, articles per request, requests per second, a hard stop with HTTP 403 and no overage, a trial before the card is billed, and the article fields needed to store each story once.
- What got in the way
- Publication timestamps were ambiguous enough that the client had to allow a missing trailing Z. Quota, article quality, and the 403-at-cap behavior were taken from the docs and never observed.
