Vercel AI Gateway adds DeepSeek Search and Fetch tools

DeepSeek Search returns relevant pages with titles, URLs and summaries, while DeepSeek Fetch retrieves page content as Markdown for a model to read. Both can be passed in the same tools object, and no separate DeepSeek account is needed. Developers can keep the same tools when switching model providers.

The helpers ship in AI SDK 6.0.321 and later on v6, or 7.0.160 and later on v7, and are enabled by updating the ai package and setting AI_GATEWAY_API_KEY. The changelog shows a setup where the model is instructed to try deepseek_search, then exa_search, then parallel_search on errors, and then use deepseek_fetch on a result URL. Vercel notes that this fallback is model-directed, so provider order and retry limits are best effort.

Failed requests, timeouts and empty results carry no tool charge. Configuration options and pricing details are in the DeepSeek Search and DeepSeek Fetch documentation.

Agents that need web search and page reading can get both from one gateway key instead of managing a separate DeepSeek account.

Because the search fallback is chosen by the model rather than enforced by the gateway, developers who need strict provider order or retry limits should test their setup and not rely on it.