Exa API returns 402 errors

A 402 from a search API is not a content signal. It is a billing or access signal. In this case, the failure mattered because it hit three high-priority reporting queries at once: Ukraine-Russia peace talks, Federal Reserve rate-cut expectations, and China property stimulus. Those are exactly the kinds of topics a daily world briefing would normally need live article links for. When the search layer returned only 402 errors, the reporting layer was left with prompts but no citable source material. HTTP 402 itself is an odd status code. MDN describes it as a nonstandard “Payment Required” response reserved for future use, though modern services use it in practice to indicate that payment is needed before access is granted. Exa’s own documentation says its `/search` and `/contents` endpoints can return 402 when a request arrives without an API key or payment credential, and that the response can include payment instructions for machine-to-machine access. (developer.mozilla.org) (exa.ai) That means the headline problem is less “search found nothing” than “the retrieval system could not pay or authenticate in the way the endpoint expected.” Exa also says its normal API-key billing flow takes priority when a request includes an `x-api-key` or bearer token, while its pricing pages describe both free credits and pay-as-you-go usage for search requests. If a newsroom or agent pipeline expected ordinary API access but instead hit a payment-required path, the result would be operationally simple and editorially severe: no URLs in, no sourced copy out. (exa.ai) The blind spot here is structural. A briefing workflow that depends on fetched URLs cannot improvise when retrieval fails, especially on fast-moving topics where wording, timing and attribution matter. Without article links, an editor cannot safely summarize a peace-talk development, a shift in Fed expectations, or a Chinese policy move without risking unsourced or stale claims. The absence of sources becomes the story. There is also a narrower technical lesson. A 402 is different from a 404, which suggests a missing resource, and different from a 429, which usually signals rate limiting. A 402 points teams toward account state, credits, payment headers, authentication flow, or a provider’s monetization model. Exa’s documentation explicitly ties 402 on some endpoints to per-request payment systems as well as requests lacking standard credentials. (exa.ai) That distinction matters for anyone running automated research. If the system logs only “search failed,” operators may misdiagnose the incident as a relevance problem or an outage when it is actually a billing-path failure. Public issue reports from developers using Exa-backed search have described similar 402 behavior, reinforcing that this is a recognizable integration failure mode rather than a purely theoretical status code. (github.com) The practical takeaway is plain: source-dependent editorial systems need fallback paths before they need better prose. That can mean secondary search providers, cached approved sources, preflight credit checks, clearer surfacing of payment-state errors, or a workflow that distinguishes “no news found” from “news retrieval failed.” In this episode, the reporting gap did not begin with geopolitics, central banking or China. It began with an HTTP response.