Free guide
DeepSeek's App Features Are Now in Its API: Web Search, Link Reading, Files and Safety
The search, link reading and file parsing behind DeepSeek's own app are now a few parameters in its API. What changed, how it works, what it costs, and how to try it today.

The short version
- Add deepseek_web_search or deepseek_read_url to the tools in your request and DeepSeek runs the searching and reading on its own servers. Your code just shows the results.
- With thinking on, search becomes deep thinking search: the model searches, opens pages and finds text in them, up to about 20 tool calls before it answers.
- It works on deepseek-flash only. DeepSeek says public-facing apps must turn on the safety check, and moderation is still your job.
How it actually works
A chef that now comes with a kitchen
With normal tool calling, the model is a chef with no kitchen. It says "search this" and your code has to run the search, open the pages and hand the results back. With these built-in tools, DeepSeek's docs say searching and fetching run on the service and your client does not execute any tool. Your app shows the blocks that come back and passes them back unchanged in the next round.
Web search vs deep thinking search
With thinking off and deepseek_web_search declared, the model decides whether to search and answers with citations, like web search in DeepSeek Chat. With thinking on, the same tool works as deep thinking search: over several reasoning steps it searches, opens pages and finds text in pages, making up to about 20 tool calls before it answers.
Links, files and safety
deepseek_read_url finds the links in the user's message, fetches those pages and answers with them in context. File input parses PDFs, Word, PowerPoint and images into the conversation. The safety check reviews input and output and returns a refusal marker when it finds a risk. There's also a conversation-title option and a user metadata marker for date, location and language.
Three API formats, one model
The chat features work in the Messages API (Anthropic format), the Responses API and the Chat Completions API, all on the deepseek-flash model. Search, fetch, OCR, speech-to-text and text-to-speech are also available as standalone APIs if you want to call them from your own code.
How to set it up
1.Get an API key
Create a key on the DeepSeek Platform (platform.deepseek.com, API keys) and top up a balance. Then put it in your shell.
export DEEPSEEK_API_KEY="sk-..."2.Try it without writing code
DeepSeek's open-source demo chat client (MIT) runs in your browser. Paste your key to try web search, deep search, URL reading, uploads, safety review and voice. The docs say your key, conversations and images are stored only in your browser. It can also import the chat history you exported from DeepSeek Chat.
https://cdn.deepseek.com/chat-demo https://github.com/deepseek-ai/chat-demo3.Turn on web search and link reading in one request
This is DeepSeek's own example in the Chat Completions format. Thinking is on, so search runs as deep thinking search. Don't name these tools in tool_choice; that returns a 400 error, because the model decides when to search.
curl https://api.deepseek.com/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer ${DEEPSEEK_API_KEY}" \ -d '{ "model": "deepseek-flash", "thinking": {"type": "enabled"}, "tools": [ {"type": "deepseek_web_search"}, {"type": "deepseek_read_url"} ], "safety_check": {"level": "high"}, "title": {"enabled": true}, "messages": [{"role": "user", "content": "What is new in DeepSeek V4.1?"}], "stream": true }'4.Show what comes back
Render the thinking block as a collapsible panel, each search step as result cards (title, site name, icon), each read page as a link, and the text with its citation markers. If stop_reason is refusal, replace the answer with your own refusal message. Pass the returned blocks back unchanged in the next turn.
5.Going public? Turn on the safety check
Keep safety_check set to level high in every request. DeepSeek says public-facing apps must enable it, and that you're still responsible for content moderation, model registration and AI-content labeling where they apply.
The fine print
- Only on deepseek-flash (DeepSeek-V4.1-Flash-1017), not deepseek-v4-pro.
- Searches cost $0.75 per 1,000 and web fetches $0.75 per 1,000, on top of token charges. A search call with three queries counts as three; failed or empty searches aren't billed.
- File parsing and OCR cost $0.225 per 1,000 pages. Text-to-speech is $0.45 per 10,000 characters and speech-to-text $0.116 per hour, both at 50% off for a limited time.
- deepseek-flash tokens: $0.30 per 1M input (cache miss) and $1.20 per 1M output at peak; off-peak is half price.
- When you turn these tools on, DeepSeek sets some parameters itself (like temperature and top_p) and ignores the values you send.
- The safety check is basic risk control, not full moderation. The demo app is for demonstration only and still uses your paid API key.
FAQ
Does the DeepSeek API have web search now?
Yes. Since November 13, 2026, adding the deepseek_web_search tool to a request on the deepseek-flash model lets the model search the web and answer with citations. DeepSeek's servers run the search, not your code.
What is deep thinking search in the DeepSeek API?
It's deepseek_web_search with thinking turned on. Over several reasoning steps the model searches, opens pages and finds text in them, making up to about 20 tool calls before it answers.
How much does DeepSeek API web search cost?
$0.75 per 1,000 searches and $0.75 per 1,000 page fetches, billed on top of normal token charges. Each query in a search call counts as one search.
Can the DeepSeek API read a link I paste?
Yes. With the deepseek_read_url tool, the model finds links in the user's message, fetches those pages and answers with their content in context. It's billed per page fetched successfully.
Does it work with the Anthropic or OpenAI SDK format?
Yes. The chat features are available in the Messages API (Anthropic format, at api.deepseek.com/anthropic), the Responses API and the Chat Completions API.
Sources
- DeepSeek's announcement on X
- Build a Chatbot (DeepSeek API docs)
- Web Tools (DeepSeek API docs)
- Models & Pricing
- Demo chat client source (MIT)
Facts checked 2026-11-14 against DeepSeek's X thread of Nov 13, the Build a Chatbot, Web Tools and Models & Pricing pages of the DeepSeek API docs, and the GitHub API for deepseek-ai/chat-demo (MIT, 1,100 stars).
Want help building with tools like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolMore free guides
How ChatGPT Wallet Pays Without Seeing Your Card (and How to Set It Up)
OpenAI's new ChatGPT Wallet lets your dot pay for things on your behalf, and neither the agent nor the store ever sees your real card number. Here's how the stand-in card works and how to set it up.
Claude Code Custom Themes: Animated Backdrops and Your Own Colors
The Claude Code desktop app can now wear a theme: your own colors and an animated scene behind the chat. Here is how a theme works, two ways to get one today, and why you should treat a theme like code.
Astryx: Give Your AI Agent Meta's Design System So Every Screen Matches
Meta's open-source design system ships 150+ React components, seven themes and a CLI your coding agent reads from. What it is, how it works, how to set it up today, and the fine print.