Overview
The Claude API lets you embed Anthropic’s state-of-the-art Claude models (Opus, Sonnet, Haiku — and the latest Claude 4 generation) into any product through simple HTTP-JSON calls.
Key capabilities
- Chat & Vision – send mixed text-and-image messages and receive multimodal answers.
- Tool / Function Calling – define JSON-schema tools so Claude can invoke your functions programmatically.
- Extended & Interleaved Thinking – stream Claude’s step-by-step reasoning when you need deeper insight, or keep responses terse for speed.
- Files & Web Search – upload documents for grounded answers or let Claude query the live web.
- OpenAI-compatible endpoint – swap base-URL, model name and key to reuse existing OpenAI integrations.
Core endpoints
| Endpoint | Purpose |
|---|---|
/v1/messages | Conversational & vision chat |
/v1/message_batches | Async bulk jobs |
/v1/files | File upload / retrieval |
/v1/models | Model catalogue & pricing |
All requests require an anthropic-version header (e.g. 2023-06-01) and an x-api-key.
Recent additions
- May 2025 – Claude 4 models, Files API, code-execution tool.
- Feb 2025 – OpenAI-compatible endpoint, vision URL sources.
- May 2025 – Web search tool, fine-grained tool streaming.
Typical use-cases
- Conversational assistants & support bots
- Autonomous coding and data-analysis with Claude Code
- Document Q&A, summarization & RAG pipelines
- Agentic RPA workflows via Computer Use beta
Anthropic enforces strict safety policies (Constitutional AI) and rates Claude 4 Opus at AI Safety Level 3.
For full pricing, rate-limits and SDKs (Python, JS/TS, Go, Java, Ruby) see the official docs.
