TinyFish Makes Web Search Free for AI Agents — But Token Costs Remain

TinyFish has announced that its web search capability for AI agents will be offered at no direct monetary cost, removing the per-query pricing barrier that has made large-scale agentic web access expensive. However, the piece highlights that the real cost of AI agents reading the web lies not in search API fees but in the token consumption required to process retrieved content — a cost that scales rapidly with the number of pages an agent reads per task. This reframing is important for developers designing agentic systems: even with free search, poorly designed retrieval loops or overly broad queries can make web-browsing agents prohibitively expensive at inference time. TinyFish's approach appears to encourage developers to think carefully about retrieval efficiency and context window management as primary engineering constraints. For anyone building agents that browse, summarize, or extract from web content, this is a practical prompt to audit token usage patterns in retrieval pipelines.