xAI starts billing X Search per post fetched, parents and quoted posts included

Price the two schedules against each other and the crossover sits at one post per call. A tool call under the current rate costs $0.005 ^[3][1]. A post under the new rate also costs $0.005, and a profile costs $0.01 ^[4][1]. A search that returns ten posts therefore costs ten times what the same search costs today, and a search that returns exactly one costs the same ^[2].

The quantity is not yours to set at call time. xAI's documentation says X Search covers keyword search, semantic search, user search and thread retrieval, and that agents decide how to use those tools during a request ^[6]. xAI also says every post returned by a search or thread fetch counts, including parent and quoted posts, with user profiles billed separately ^[5]. One thread fetch on a deep reply chain bills for the chain, and each additional thread the agent expands adds to that bill without a separate approval step.

RuntimeWire reports that Robert Scoble has repeatedly said publicly that Aligned News pulls roughly 30,000 X posts per day and spends about $100 to $150 per day on X API calls ^[9]. At $0.005 per post, that volume is $150 per day, about $4,500 per month, for post retrieval alone ^[10]. Against the bottom of his stated range that is 1.5x ^[4]. Add 3,000 profile fetches a day and the retrieval layer runs around $5,400 per month before Grok inference ^[11]. Aligned News says it monitors 63 curated X lists covering more than 100,000 accounts and retrieves more than 5,000 posts per sweep ^[8].

The near-parity result is the number to check before borrowing it. RuntimeWire's 37,680 posts and 1,560 profiles from August 11 through September 10 price out at $188.40 and $15.60, so $204.00, against an actual bill of $206.34 for the same period ^[12][13]. That is $2.34 lower, about 1.1 percent ^[5]. For a $206.34 bill to have been per-call at $0.005, it would have covered about 41,268 calls, which works out to 0.91 posts returned per call across the month ^[6]. Keyword and semantic search do not usually return under one post per call. RuntimeWire gives the bill without saying which rate schedule produced it ^[17].

What the two systems differ on is retrieval volume for the same job: roughly 30,000 posts a day against roughly 1,256, about 24x ^[14]. RuntimeWire's author wrote that he "never wanted the newsroom to operate by continuously vacuuming up social content and asking a model to decide what mattered afterward" ^[15]. That preference used to be a matter of design, before per-item pricing turned it into a line on the invoice under the new schedule.

In my context the control that follows is a retrieval budget per request, enforced where tool results come back rather than in the prompt. Count returned items before they reach the model, cap posts per request and cap thread expansion depth, then fail the next search when the budget is spent. Alarm on daily item counts, not on daily dollars, because item counts move the same day the agent's behavior does. RuntimeWire notes that any real bill still depends on architecture, caching, deduplication and how much surrounding context gets fetched ^[17].