DeepSeek opens DeepSeek App search, file and voice tools via API

DeepSeek opened the server-side conversation tools powering the consumer DeepSeek App through the DeepSeek API on 13 November 2026, paired with an MIT-licensed browser frontend at github.com/deepseek-ai/chat-demo. Following DeepSeek Harness Creator Mode, developers can call these built-in tools across three protocol endpoints using the model ID deepseek-flash: OpenAI-style Chat Completions (api.deepseek.com/chat/completions), the Responses API (api.deepseek.com/responses) and an Anthropic-compatible Messages endpoint (api.deepseek.com/anthropic/v1/messages).

Server-side search, URL fetching, document OCR and speech endpoints

Enabling deepseek_web_search runs either standard cited web and image search or multi-step Deep Thinking Search when reasoning mode is active, while deepseek_read_url fetches URLs pasted into user prompts directly on DeepSeek’s servers. Client applications render the returned thinking, server_tool_use, tool-result and citation blocks and pass those blocks back unchanged on subsequent turns.

DeepSeek API assistant message diagram ordering thinking, server_tool_use with web_search_tool_result, web_fetch_tool_result and cited text blocks — DeepSeek’s API streams interleaved thinking, server-side search and URL-fetch result blocks ahead of cited text.Image: DeepSeek

Additional request fields and standalone endpoints cover PDF, Word, PowerPoint and image parsing, optical character recognition (OCR), speech-to-text (ASR), text-to-speech (TTS), automatic conversation title generation and input-output content safety checks. DeepSeek’s chatbot guide did not publish separate per-call fee schedules for the built-in search or parsing tools.

Key details

Supported API protocols

Chat Completions (/chat/completions), Responses (/responses) and Anthropic Messages (/anthropic/v1/messages)

Reference model and client

deepseek-flash model ID; MIT-licensed static web app at github.com/deepseek-ai/chat-demo (Node.js 22.12+)

Server-executed tools

deepseek_web_search (standard and Deep Thinking Search), deepseek_read_url, file parsing, OCR, ASR and TTS

Client storage design

API key stored in browser localStorage; conversation trees and attachments persisted in IndexedDB

Release details & context

Sampling overrides under content safety and static browser deployment

DeepSeek’s documentation notes a key runtime constraint for production deployments: turning on the built-in input and output content safety filter automatically overrides custom temperature and top_p sampling parameters on that request.

The companion chat-demo frontend compiles to static assets hosted at cdn.deepseek.com/chat-demo and GitHub Pages with zero backend server required, supporting KaTeX maths rendering, Mermaid diagrams, conversation branch switching, Simplified Chinese and English interfaces, and direct import of ZIP chat-history archives exported from the consumer DeepSeek Chat service.

Common questions

Do client apps need to execute web scraping locally when calling deepseek_web_search or deepseek_read_url?

No, DeepSeek executes both search and URL fetching on its own servers and returns structured result and citation blocks that the client simply passes back on multi-turn requests.

What happens to custom temperature and top_p settings when DeepSeek’s content safety check is enabled?

Enabling the API’s built-in input-output content safety parameter overrides custom temperature and top_p values on those calls.

DeepSeek release timeline (4 launches tracked)

Earlier17 Oct 2026

DeepSeek releases DeepSeek-V4.1-Flash-1017 with MIT-licensed open weights

DeepSeek says DeepSeek-V4.1-Flash-1017 scores 77.0 on DeepSWE v1.1, ahead of Claude Opus 5’s 74.0 and GPT-5.6 Sol’s 73.0.

Earlier20 Oct 2026

DeepSeek open-sources DeepInfer, the stack behind V4.1-Flash

DeepSeek’s DeepInfer is the open-source inference code that serves DeepSeek-V4.1-Flash in production.

Earlier2 Nov 2026

DeepSeek Harness adds Creator Mode for building live UI plugins in chat

DeepSeek added Creator Mode to its open-source DeepSeek Harness so users can build and hot-install custom UI plugins and themes from chat.