Subtotal $0.00
Email : 1
The Claude API lets developers build directly on top of Anthropic’s models instead of using the Claude.ai chat interface. Interest in it has grown sharply this year. In fact, search demand for the Claude API has more than tripled since early 2025. This guide covers what the Claude API actually is. Specifically, it walks through current 2026 pricing for every model. Finally, it shows you how to get an API key and make your first call, step by step.
Quick Answer
The Claude API is Anthropic’s developer platform for building custom applications with Claude models. You sign up at platform.claude.com and generate an API key. From there, you call the Messages API from Python, TypeScript, or several other supported languages. Pricing starts at $1 per million input tokens on Haiku 4.5. It goes up to $10 per million input tokens on Fable 5. Output tokens cost roughly five times more than input across every model.
What Is the Claude API?
The Claude API is a set of endpoints that let your own software talk directly to Claude models. Instead of typing into a chat window, your code sends a request and gets a structured response back. Developers use it to build chatbots, coding assistants, research tools, and autonomous agents. Unlike the Claude.ai subscription, the API bills you per token rather than charging a flat monthly fee. That makes it a better fit for products with unpredictable or high-volume usage. Our Claude Sonnet 5 vs GPT-6 Astra comparison looks at how Claude’s models stack up against rival AI systems in practice.
Claude API Pricing in 2026
Anthropic prices the Claude API per million tokens, and rates vary significantly by model. Haiku 4.5 is the cheapest option, at $1 per million input tokens and $5 per million output tokens. Sonnet 5 costs $2 per million input tokens and $10 per million output tokens. That rate became permanent in August 2026. Opus 5 runs $5 per million input tokens and $25 per million output tokens. Fable 5, Anthropic’s most capable model, costs $10 per million input tokens and $50 per million output tokens. Across every model, output tokens cost about five times more than input tokens.
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Haiku 4.5 | $1.00 | $5.00 |
| Sonnet 5 | $2.00 | $10.00 |
| Opus 5 | $5.00 | $25.00 |
| Fable 5 | $10.00 | $50.00 |
Two discounts can meaningfully lower your bill. Prompt caching charges just 10% of the base input price for cached reads. As a result, repeated context pays for itself after a single reuse. The Batch API cuts both input and output rates in half for requests that can run asynchronously within 24 hours. Under batch pricing, for example, Opus 5 drops to $2.50 per million input tokens and $12.50 per million output tokens.
How to Get Your API Key
Getting started takes only a few minutes. First, create an account at platform.claude.com. Second, open the API keys section of your account settings, as Anthropic’s official documentation walks through in detail. Third, generate a new key and store it somewhere secure, since Anthropic only shows it to you once. Finally, set that key as an environment variable so your code can read it without hardcoding it. On most systems, that looks like the command below. Keep this key private. Anyone who has it can make API calls billed to your account.
export ANTHROPIC_API_KEY="your-api-key-here"Making Your First Claude API Call
Once your key is set, install the official Python SDK using the command below. Anthropic also offers SDKs for TypeScript, Go, Java, C#, Ruby, and PHP. That means you can work in whichever language your project already uses.
pip install anthropicHere’s a minimal example that sends a prompt and prints Claude’s reply:
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-sonnet-5",
max_tokens=1000,
messages=[{"role": "user", "content": "Explain what the Claude API does in one paragraph."}]
)
for block in message.content:
if block.type == "text":
print(block.text)Running that script sends your prompt to Claude. It then prints the response to your terminal. From there, most developers move on to multi-turn conversations, system prompts, and tool use. The Messages API supports all of these natively.
API vs. the Claude.ai Subscription
Choosing between the two comes down to how you plan to use Claude. The Claude.ai subscription makes sense for individual work: writing, research, and everyday chat, all for a flat monthly fee. However, the API makes better sense for a different kind of project. Specifically, it fits when you’re building a product, an internal tool, or an automated workflow that calls Claude programmatically. Many teams use both. For instance, they rely on Claude.ai for personal productivity while building customer-facing features on the API. Our Claude vs GPT-6 Astra breakdown covers pricing and performance side by side, if you’re weighing Claude against other frontier models for a build decision.
Key API Features Worth Knowing
A few features set the current Claude API apart from a basic chat endpoint. Tool use lets Claude call your own functions, such as a database lookup or a calculator, mid-conversation. Extended thinking gives Claude extra reasoning steps before it answers, which helps on harder problems. Vision support means you can send images alongside text in the same request. The full 1 million token context window is also available, with no long-context surcharge. As a result, processing large documents doesn’t cost extra per token. These same capabilities power the agent-based systems covered in our guide to Claude managed agents and agentic workflows. For the latest model announcements and version history, our Claude AI latest update page tracks changes as Anthropic ships them.
Frequently Asked Questions
Is the Claude API free to use?
No, but new accounts typically receive a small amount of free credit to test the API. That credit lets you try things out before adding a payment method. Beyond that trial credit, you pay per token based on the model you use.
Which Claude model is cheapest for the API?
Haiku 4.5 is the least expensive option, at $1 per million input tokens and $5 per million output tokens. It works well for simple, high-volume tasks like classification or short replies.
Do I need to know how to code to use the Claude API?
Yes, the API is designed for developers. If you just want to chat with Claude without writing code, the Claude.ai website or app is the better choice instead.
What programming languages are supported?
Anthropic provides official SDKs for Python, TypeScript, Go, Java, C#, Ruby, and PHP. You can also call the API directly over HTTP from any language using cURL.
Final Thoughts
The Claude API gives developers direct, programmatic access to Anthropic’s models. Pricing scales from $1 per million tokens on Haiku 4.5 up to $50 per million output tokens on Fable 5. Getting started only takes a few steps: create an account, generate a key, and send your first request. From there, prompt caching and batch processing can cut your costs substantially as usage grows. The API is the right tool once you need Claude to run inside your own software rather than inside a chat window.
