Claude API for beginners: pricing, models and your first call
The Claude API lets developers send text, documents and images to Claude models and receive responses in their own apps. In October 2026 the main models are Haiku 5.5 ($0.10 input and $0.50 output per million tokens for prompts up to 100,000 tokens), Sonnet 5.5 ($2 and $10), Opus 5.5 ($4 and $20) and Fable 5.1 ($10 and $50). You create an API key in the Claude Console, send a request to the Messages API and pay per token used.
What is the Claude API?
The Claude API is Anthropic’s developer interface to its models. Instead of chatting in the Claude app, your software sends a request with a prompt, and optionally documents, images or tool definitions, and receives Claude’s response. Businesses use it to add AI features to products, automate document processing, power support tools and build agents. It is also available through Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry.
Which model should you choose?
| Model | Use it for |
|---|---|
| Haiku 5.5 | High-volume, fast tasks: classification, extraction, routing |
| Sonnet 5.5 | The everyday default: chat, drafting, analysis, most coding |
| Opus 5.5 | Long-running agentic coding and complex knowledge work |
| Fable 5.1 | The most demanding reasoning and long-horizon agent work |
Prices are from Anthropic’s pricing page as summarised in October 2026 by DEV Community. Our comparison of Opus, Sonnet and Haiku and our guide to Fable and Mythos explain the trade-offs.
What is a token?
A token is a small chunk of text, often part of a word. You pay for input tokens (what you send) and output tokens (what Claude writes back). Output costs five times as much as input on current models, so asking for concise answers saves money. Long documents add input tokens on every request unless you use prompt caching.
How do you make your first call?
- Create an account and key: Sign up for the Claude Console, add billing and create an API key. Keep it secret.
- Try the Workbench: Test prompts in the Console before writing code.
- Install an SDK: Use the official Python or TypeScript SDK, or call the HTTP endpoint directly.
- Send a message: Choose a model, set a maximum number of output tokens and send a user message.
- Read and log the response: Store the answer and the token usage reported with it.
- Set limits: Configure spend limits and alerts in the Console.
A first request contains just three things: the model name, a maximum output length and a list of messages. Everything else, such as system instructions, tools and images, can be added later.
How do you keep costs down?
- Pick the smallest model that works: Test Haiku 5.5 or Sonnet 5.5 before reaching for larger models.
- Prompt caching: Reuse long, repeated context; cache reads cost a fraction of normal input.
- Batch API: Half price for jobs that can wait for results.
- Shorter outputs: Ask for concise formats and set sensible output limits.
What else can the API do?
- Tool use: let Claude call your functions and APIs
- Vision: read images, charts and scanned documents
- PDF support: send documents directly
- Long context: up to a million tokens on current models
- Structured output: ask for JSON that matches a schema
- MCP: connect to tools through the Model Context Protocol
How do you estimate monthly cost?
Multiply your expected monthly input tokens by the model’s input price and your output tokens by its output price, both per million. A simple chatbot handling 10 million input tokens and 2 million output tokens a month would cost about $40 on Sonnet 5.5 and around $2 on Haiku 5.5 before caching, according to worked examples by Growthcab. Agent workloads use far more tokens per task, because each step resends context, so test with realistic sessions and use caching from the start.
What should a first project be?
Pick a narrow, internal task where mistakes are cheap: summarising support tickets, tagging customer feedback or extracting fields from invoices. Build a small evaluation set of 20 to 50 real examples with known correct answers, and test each prompt and model against it. That habit, more than any single prompt trick, is what separates reliable AI features from demos.
What should you get right before launch?
- Never put your API key in front-end code
- Validate and limit user input
- Set spend limits and monitor usage
- Log prompts and responses with care for personal data
- Test with realistic and awkward inputs
- Have a fallback if the API is slow or unavailable
To compare providers, read OpenAI, Claude and Gemini APIs compared; to keep bills down, see how to cut LLM API costs.
How UIUX Studio helps
Our product engineering team builds AI features on the Claude API, from model choice and prompt design to cost controls and monitoring. If you prefer to work in a coding agent, see our Claude Code beginner’s guide.
Frequently asked questions
Is the Claude API free?
No. It is pay-as-you-go per token, though new accounts may receive trial credit.
Which model is cheapest?
Haiku 5.5, at $0.10 input and $0.50 output per million tokens for prompts up to 100,000 tokens.
Is a Claude subscription the same as API access?
No. Pro and Max plans cover the Claude apps; the API is billed separately per token.
Which languages are supported?
Official SDKs exist for Python and TypeScript, and any language can call the HTTP API.
Can I use the API through AWS or Google Cloud?
Yes. Claude models are available on Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry.