Exa API blocks three queries
- Exa’s overnight prefetching failed on September 2, 2026, when three news searches returned HTTP 402 responses and left the briefing bundle without usable source material. - The key detail was the repeated “402 Payment Required” response, which Exa’s documentation says can be returned by `/search` and `/contents`. - The next step is operational: rerun the fetch with valid billing or paste article URLs into the workflow.
Exa’s overnight prefetching run failed on three separate news searches, returning HTTP 402 responses instead of article text, excerpts or source links. The missed queries were “Ukraine drone strikes Russia,” “China property stimulus measures,” and “Nvidia antitrust probe AI,” according to the source pack. That left the briefing bundle without the underlying reporting needed to produce a sourced world-news roundup. Exa’s own documentation says its `/search` and `/contents` endpoints can return 402 when a request arrives without an API key or payment header. ### Why did three unrelated news topics all fail the same way? The common failure was not editorial but technical: each query produced the same HTTP status code. MDN describes 402 Payment Required as a nonstandard client-error response code reserved for future use and associated with payment systems rather than a content-specific problem. That makes the repeated 402 responses look like an access or billing issue in the retrieval layer, not three separate reporting gaps tied to Ukraine, China or Nvidia. (exa.ai) Exa’s documentation gives two supported paths for paid access. The company says requests can be billed through a standard API-key flow tied to account credits, or through per-request payment headers on supported endpoints. If neither is present, Exa says `/search` and `/contents` can answer with 402 and a payment challenge header. ### What does a 402 response mean in Exa’s system? Exa says 402 is part of its payment-handling design, not a generic outage code. (developer.mozilla.org) In its x402 guide, the company says a request sent without an API key or payment header can trigger 402 on `/search` and `/contents`, with pricing details returned in the `PAYMENT-REQUIRED` header. Its billing documentation separately says the API runs on a pay-as-you-go credit system, with charges based on usage and balances managed in the dashboard. (exa.ai) The pricing documents show why a billing break can stop a prefetch job immediately. Exa says Search is billed per 1,000 requests and Contents per 1,000 pages, while free accounts receive a limited monthly credit allotment. If a workflow depends on those endpoints to gather article text before writing begins, a payment challenge at fetch time means the downstream bundle never gets populated. (exa.ai) ### Why did this block the briefing instead of just degrading it? The source pack says the failed searches left no article text, no excerpts and no URLs to cite. That matters because the writing workflow required source-backed reporting rather than unsourced summaries or recalled facts. Without links or extracted text, the system had no material to quote, paraphrase or attribute for any of the three topics. The result was a hard stop, not a thinner draft. (exa.ai) The pack’s own recommendation was operational: rerun the fetch or paste article URLs directly so the contents step can retrieve material for the briefing. ### Could this have been a rate-limit problem instead? Exa publishes separate documentation for rate limits and says most endpoints are limited by queries per second. But rate-limit handling is documented separately from payment handling, and the source pack identified 402 specifically, not a throttling code. On the available evidence, the failure points to payment or authorization setup rather than request volume. ### What should happen next in the workflow? The immediate fix is to rerun the failed searches with working billing credentials or a valid payment mechanism attached to the request. Exa says API-key billing takes priority when an `x-api-key` or bearer authorization header is present, while its alternative payment flow relies on the payment header challenge-response process. (exa.ai) If the automated fetch still fails, the fallback is manual. Pasting article URLs into the workflow would let the contents-retrieval step attempt extraction directly, and the next successful bundle would need article text, excerpts and source links for the three named topics before a sourced world briefing can be produced. (exa.ai)