DeepSeek Search and Fetch API: Pricing, Free in Claude Code and Codex Until Nov 1

If you use the DeepSeek API inside Claude Code, Codex or DeepSeek Harness, your web searches are about to cost money. DeepSeek added two new web tools, Search and Fetch, to its API this week, and its docs say the calls those coding tools already make to them are free only until 00:00 Beijing time on November 1, 2026.

The Search API runs a query against DeepSeek's own search engine and returns structured results, and the Fetch API takes a single URL and returns the page as Markdown. According to DeepSeek's documentation, both are aimed at agents and RAG applications that need live web content, and both use the same API key as the rest of the DeepSeek API.

Pricing and the free period

If you call either API directly, you pay from day one: $0.75 per 1,000 successful requests (CNY 5 in DeepSeek's Chinese docs). DeepSeek says a request is only billed when it returns HTTP 200 with a result, so failed calls, timeouts and searches that come back empty are free. The number of results you ask for doesn't change the price.

The free window only covers calls the coding tools make on their own. The docs say that when the DeepSeek API is used through DeepSeek Harness, Claude Code or Codex, server-side tool calls already go through Search and Fetch, and that those Search and Fetch calls switch to per-request billing on November 1.

What the two APIs return

A Search request takes a query of up to 128 characters and can return between 1 and 50 results, with 10 as the default. Search syntax in the query is passed through as is, so operators like site:, -site:, exact-phrase quotes, lang:, after: and before: all work. Each result comes with a title, URL, a short snippet for display, a longer summary of typically 600 to 800 characters meant for a model to read, and optionally the full page content. DeepSeek also attaches a relevance score from 0 to 3 and an authority tier from 0 to 4 to every result.

Fetch is simpler. You send one HTTP or HTTPS URL and get back the page title and its content as Markdown, subject to the target site's own availability and access rules.

Rate limits

Each account gets 5 requests per second for Search and another 5 for Fetch, counted separately and regardless of which API key is used. These limits are also separate from model concurrency limits. Fetch has an additional per-domain limit, and going over either limit returns HTTP 429.