DeepSeek API Docs

The DeepSeek API gives developers programmatic access to DeepSeek’s large language models. You send chat completion requests using the familiar OpenAI or Anthropic API formats, and get responses from models like deepseek-v4-flash or deepseek-v4-pro. The API supports both streaming and non-streaming modes, so you can use it for real-time chatbots or batch processing. DeepSeek even includes a dedicated Anthropic-compatible endpoint, making it easy to switch without rewriting code.

What makes the DeepSeek API stand out is its compatibility layer. You can plug it into existing tools and SDKs that already work with OpenAI or Anthropic — just change the base URL and API key. The page highlights integrations with popular agent tools like Claude Code, GitHub Copilot, and OpenCode, so you can run those assistants on DeepSeek's backend instead. You also get configurable thinking modes and reasoning effort parameters, which let you trade off speed for depth depending on your use case. The API supports system messages, multi-turn conversations, and streaming responses out of the box.

This API is best suited for developers and teams who want to experiment with or deploy DeepSeek’s models without learning a new interface. If you already have an app using OpenAI’s API, migration is straightforward. It’s also useful for AI engineers building custom chat agents or coding assistants who want a cost-effective alternative. The documentation is lean but clear, and the agent integration guide helps you get up and running quickly.