DeepSeek Launches Search and Fetch APIs for Live Web Integration

DeepSeek has introduced Search and Fetch APIs to enable AI applications to access and process live web content. The Search API provides structured results via an in-house engine, while the Fetch API returns page content in Markdown format. These services are billed at $0.75 per 1,000 successful requests and feature a 5 QPS rate limit.

Key Points

• DeepSeek's Search and Fetch APIs provide structured, real-time web access for AI agents and RAG systems.
• The Search API supports advanced operators like 'site:', 'lang:', and date filters to refine results from an in-house engine.
• The Fetch API converts web page content directly into Markdown for seamless integration with LLM prompts.
• Billing is simplified to a flat rate of $0.75 per 1,000 successful requests, with no charges for failed or empty results.
• Both services are subject to a rate limit of 5 QPS per account and use standard API key authentication.

Sentiment

Mixed; users are impressed by the technical offering and price point but deeply divided on the company's ethics and the impact of AI scraping on the web.

In Agreement

• DeepSeek's pricing is highly competitive and significantly undercuts established players like Brave and Kagi.
• Hosted search APIs are essential for production AI agents because self-hosting search (e.g., SearXNG) is unreliable due to aggressive rate-limiting.
• Using a search API from the same provider as the LLM is more convenient and potentially more optimized for the model's training.
• The Fetch API's ability to return Markdown is highly beneficial for RAG and AI agent processing.

Opposed

• DeepSeek is viewed by some as a 'bad actor' due to allegations of routing user data to third-party models like Claude and using fraudulent payment methods.
• Data retention policies are a concern, as DeepSeek logs search history and target URLs, which is a dealbreaker for privacy-conscious users.
• The aggressive scraping required to power these indices is contributing to an 'arms race' that makes the internet harder for human users to navigate.
• Some users argue that higher-priced providers like Exa offer superior semantic search quality and better signal-to-noise ratios compared to standard keyword search.