DeepSeek Releases Search and Fetch Web Data APIs for Developers
DeepSeek launched Search and Fetch APIs on October 17, 2026, priced at $0.75 per 1,000 successful calls each, with promotional free search for select developer
When building research agents or automation pipelines, managing external search engines and maintaining web extraction scrapers often introduces considerable architectural friction. On October 17, 2026, DeepSeek officially released dedicated Search and Fetch APIs for software developers. Confirmed by tech observer @JungkwanDean, this release provides direct primitives to discover URLs and ingest clean page contents from the open web without maintaining separate extraction pipelines.
Dual Primitives for Web Discovery and Content Retrieval
The interface decouples data access into two distinct functional primitives: Search and Fetch. The Search function queries the web to find relevant pages, handling reconnaissance. Meanwhile, the Fetch function reads the body content of a designated page, handling extraction. This separation fits research tasks well; for example, an agent can search for competitor updates and then fetch original notice texts to generate structured comparative reports without custom scrapers.
Usage-Based Pricing and Coding Environment Promotion
Both endpoints require an API key and follow a metered billing model based on successful requests. DeepSeek prices both the Search API and Fetch API at $0.75 per 1,000 successful calls each. Because requests are billed independently, search queries and subsequent page fetches accumulate per transaction. In addition, server-side search calls made through Claude Code, Codex, and DeepSeek Harness are offered free of charge until November 1, 2026, at 01:00 KST, aiding early integration tests.
Operational Boundaries and Practical Considerations
Adopting these endpoints requires observing explicit operational boundaries. The temporary promotional waiver applies strictly to server-side search calls within the specified environments; underlying AI model token usage remains normally billable. Furthermore, this announcement introduces data collection APIs and does not involve model weight releases or open-source weights. Teams scaling up data collection must assess the $0.75 per 1,000 requests fee structure against high-volume alternatives, while enjoying streamlined tooling for interactive agents.