DeepSeek Opens App Tools to Developers via API

DeepSeek has opened the tools behind its consumer chat app to developers through its application programming interface (API), the Hangzhou-based AI lab announced on 13 November 2026. The update lets outside builders use the same capabilities that power the DeepSeek app, including web search, file reading, link reading, safety checks and voice features.
The centrepiece is a model called deepseek-flash, based on the DeepSeek-V4.1-Flash-1017 checkpoint. Developers can switch on the new built-in tools with parameters in existing requests, using the Messages, Responses and Chat Completions API formats.
What developers can now do
With the web search tool enabled, the model can search the web or images and answer with citations. With thinking mode switched on, it can search and read pages across several reasoning steps. A separate tool, deepseek_read_url, fetches the contents of links that appear in a user's message.
File input is also part of the release. Developers can upload PDF, Word and PowerPoint documents, as well as images, and the model can answer questions using their contents. Documents can be extracted as plain text or parsed in a way that preserves their layout, including tables, charts and images.
The release also adds automatic conversation titles and user-context metadata, and a content safety check that runs inside the model call.
Standalone services for search, documents and voice
Alongside the built-in tools, DeepSeek has introduced five standalone APIs: Search, Fetch, OCR, speech recognition (ASR) and text-to-speech (TTS). Search runs queries through DeepSeek's own search engine and returns structured results, while Fetch retrieves a web page and returns it in Markdown format. The OCR service turns images of printed text, including scanned documents, into machine-readable words.
The tools are billed separately from token charges. According to DeepSeek's pricing pages, as reported on 13 November, search and web fetch cost $0.75 per 1,000 successful requests, while file parsing and OCR cost $0.225 per 1,000 pages. The deepseek-flash model itself is priced at $0.30 per million uncached input tokens and $1.20 per million output tokens at peak hours, with off-peak rates at half that.
Conditions and limits
There is one notable condition. Public-facing applications built on the API are required to include safety checks, DeepSeek said. The company also released an open-source demo client alongside the expansion, so developers can see how the tools fit together.
The move follows DeepSeek's opening of standalone Search and Fetch APIs on 17 October, when it also updated the V4.1-Flash-1017 checkpoint. The latest release goes further by bringing the app's everyday features, rather than only raw model access, into developers' own products.
For builders, the practical change is straightforward: an app that already calls DeepSeek's API can now add live search, document understanding and voice without building those services itself. The cost is that each tool call adds its own fee on top of the price of the tokens the model uses.