DeepSeek’s October 17, 2026 announcement opened standalone Search and Fetch APIs for developers building agents and retrieval-augmented generation (RAG) applications. Search uses DeepSeek’s in-house search engine to find live web pages; Fetch retrieves a specified page as Markdown. Each endpoint is priced at $0.75 per 1,000 successful requests, while some server-side tool calls were announced as temporarily free.
14
23
24
Endpoints, format and authentication
Both APIs use the base URL https://api.deepseek.com and return raw JSON. The documented endpoints are:
23
- Search:
POST https://api.deepseek.com/v1/tools/search
- Fetch:
POST https://api.deepseek.com/v1/tools/fetch
The guide says the APIs use the same bearer authentication as the DeepSeek API.
23
Search inputs and result details
Search requires a nonempty query of 1–128 Unicode characters. The optional count parameter sets how many results to return: it defaults to 10 and accepts values from 1 to 50. The documented optional inputs also include ori_question, localization and need_content; an integration reference lists need_content as defaulting to false. The supplied excerpts do not establish the exact semantics or expected format for every optional input.
17
36
The official API reference identifies site:, -site: and quoted phrases as supported query syntax. An integration reference also lists -term, lang:, after: and before:; treat that additional operator list as implementation guidance rather than a complete guarantee of API behavior.
17
34
Search returns ranked, structured results. A community integration reference lists fields including title, url, snippet, summary and, when requested, content, as well as date, rerank_score and authority. It describes snippets as typically 100–200 characters and summaries as 600–800 characters. Because those field details and lengths come from an integration reference rather than the API reference excerpt, they should be verified against the current response schema before an implementation depends on them.
36
Fetch inputs and output
Fetch requires a URL of up to 8,192 characters. The API reference specifies HTTP or HTTPS URLs with a host and disallows URLs containing a username or password. Its response provides the page title and content as Markdown.
18
The available API documentation does not establish that every site can be accessed, or specify all extraction restrictions and failure cases. Do not assume Fetch will return complete content for every URL; the supplied evidence does not define a universal site-access or extraction guarantee.
18
Pricing and temporary free calls
DeepSeek announced a price of $0.75 per 1,000 calls for each endpoint. Reporting describes the charge as applying to successful requests and says failures, empty results and timeouts are not billed. An integration reference describes retrieval charges as separate from model-token charges, so the API call fee should not be treated as including any model usage needed to process retrieved content.
2
13
24
The launch announcement said Search and Fetch calls made through server-side tools in DeepSeek Harness, Claude Code and Codex were temporarily free for DeepSeek API users. It scheduled the start of billing for October 31, 2026, at 16:00 UTC—November 1 at 00:00 Beijing time. This exemption concerns those tool integrations; it should not be read as a blanket waiver for direct API calls.
2
24
Rate limits and what remains unclear
Secondary coverage reports a limit of 5 requests per second per account for the Search and Fetch APIs.
4
10 The supplied documentation does not establish whether the endpoints have independent limits, whether enforcement is per account or API key, whether Fetch has additional per-domain limits, or exactly how the API responds when a limit is exceeded.
17
18
DeepSeek’s rate-limit documentation separately describes model concurrency limits, but the available excerpt does not explain how those limits interact with Search and Fetch traffic. It also does not confirm that exceeding a Search or Fetch limit always returns HTTP 429.
25
In short, the launch established the core Search and Fetch interfaces, their in-house search backing, and request-based pricing. For implementation details beyond the documented parameters and URL requirements—especially optional-field formats, complete response schemas and edge-case behavior—check the live API references rather than assuming third-party integration notes define the service contract.
17
18
23