EN AR

DeepSeek opens web search, file reading and safety tools to developers

Khaled Aziz

Also in: Companies
DeepSeek released its full application API suite, including web search, file input, URL reading and voice.

Key Points

  1. DeepSeek released its full application API suite, including web search, file input, URL reading and voice.
  2. The features already power the DeepSeek App; developers now enable them through existing API parameters.
  3. Developers can skip building search, document parsing and content moderation layers themselves.

The latest:

DeepSeek has opened the capabilities behind its consumer app to outside developers, releasing a full suite of application APIs covering web search, file input, URL reading, model safety, speech recognition and text-to-speech. According to the company’s change log dated 13 November 2026, enabling the matching parameters in existing APIs returns these conversation features directly, with tool calls and context assembly handled on DeepSeek’s side.

Details:

  • What shipped: According to the DeepSeek API change log, the release covers web search, file input, URL reading, model safety, ASR and TTS. The company said these capabilities already work alongside model inference inside the DeepSeek App and its web version to produce the conversation experience users have today.
  • The integration model: DeepSeek said developers do not call separate services or stitch results together manually. Turning on the relevant parameters in the existing APIs delivers web search, file input and URL reading, with the tool calls and the assembly of context into the conversation handled by DeepSeek’s service.
  • Which model, which APIs: The documentation specifies that deepseek-flash supports the new features across three interfaces: the Messages API, the Responses API and Chat Completions. Other models in the DeepSeek lineup are not named as supporting the suite.
  • Web search: With thinking disabled and the deepseek_web_search tool enabled, the model runs web or image searches as the question requires and answers with citations, according to the documentation, which describes the citation behavior as part of the answer rather than an optional add-on.
  • Deep thinking search: A second mode combines thinking with the same search tool. DeepSeek said that in this configuration the model searches the web and opens links on its own across multiple reasoning steps, rather than running a single query before answering.
  • Documents and links: File input accepts PDFs, Office files and images, added to the conversation context after plain-text extraction or rich-layout parsing. Separately, the deepseek_read_url tool lets the model detect links inside a user message, fetch those pages and answer from them.
  • Safety controls: A safety check parameter reviews both model input and output and returns a refusal marker when a risk is detected, which DeepSeek described as basic risk control. The company advised anyone serving the public to enable it and to meet obligations including content safety review, large model registration and AI labeling.
  • Standalone pieces: Beyond the bundled parameters, DeepSeek said the underlying capabilities and the voice features are also offered as separate APIs, listing Search and Fetch, OCR, ASR and TTS, letting developers call individual components without routing everything through a chat interaction.
  • Open-source demo: DeepSeek said it built and open-sourced a demo chat client on top of the new APIs, and that the client can import a user’s existing DeepSeek chat history. The company pointed developers to its Build a Chatbot guide for usage and to its Models and Pricing page for costs.

Between the lines:

The safety parameter is framed as a compliance tool, not just a filter: DeepSeek ties it to content review, model registration and AI labeling duties that fall on public-facing services in China. Bundling moderation into the same API as search and file parsing lowers the cost of launching a consumer chatbot built on DeepSeek, since the regulatory layer arrives with the capability layer.

What’s next

Pricing for the new endpoints sits on DeepSeek’s Models and Pricing page. Watch whether support extends beyond deepseek-flash to other models, and how the open-sourced demo client is adopted by developers migrating existing chat histories.

What to read next