Definition
The DeepSeek Search / Fetch API is DeepSeek’s search and page-retrieval interface: the official documentation describes the Search API as finding “web pages for a query and return[ing] structured results”, powered by DeepSeek’s in-house search engine, with a companion Fetch API that retrieves a single web page and returns its content as Markdown. See the official guide.
How it works
Both endpoints are bearer-authenticated JSON APIs: POST /v1/tools/search (search) and POST /v1/tools/fetch (page retrieval). The search query takes 1–128 characters and supports operators such as site:, "...", lang:, and after:/before:; count ranges from 1–50; results carry title, URL, snippet, summary and a rerank score. Billing is per successful request — the official documentation prices both Search and Fetch at $0.75 per 1,000 successful requests, with failed requests and empty results not billed; current prices and rules follow the official guide.
Usage
It targets developers wiring external information into applications or agents: the caller receives structured results and decides how to feed them into context. The official documentation notes that when using models through DeepSeek Harness, Claude Code or Codex, server-side tool calls already hit this Search/Fetch API, and that this usage starts being billed per request from 00:00 Beijing time on November 1, 2026. It can be wrapped directly into a retrieval pipeline when Chinese-language search capability is needed.
Related and confusable terms
This is a separately billed API service, not the web-search toggle in the chat app and not the DeepSeek models themselves: any model can call the two endpoints over HTTP. Keep it distinct from the models’ text generation; its relation to built-in chat search is covered in the DeepSeek Chat entry.
Discussed on Next Token
In episode 007, in the chapter “搜索 API：给个人 Agent 补上中文信息” (Chinese transcript), 橘子 said the search API had come out that week and called it “非常好用……基本是国内最好用的 API 之一”; 向阳乔木 said he “立马把它 API 包成了一个 Skill” and found it “响应又快，质量又好”; 橘子 also noted the API was exposed as a standalone service any model can use. These are the participants’ experiences; billing and other facts follow the official documentation (full chapter).
Frequently asked questions
How is the DeepSeek Search API billed?
Per successful request: the official documentation prices both Search and Fetch at $0.75 per 1,000 successful requests, and failed requests or empty results are not billed; current pricing is on the official guide.
What do I need to use the DeepSeek Search API?
A DeepSeek API Key, created on the open platform and sent as a bearer token to POST /v1/tools/search and POST /v1/tools/fetch; integration details are in the official guide.
Is the Search API tied to DeepSeek models?
No. The two endpoints are standalone HTTP interfaces any model can call; the official documentation additionally notes that server-side tool calls in Harness, Claude Code and Codex already use this API.