How to Get a Claude API Key

A Claude API key is created in Anthropic’s developer console at platform.claude.com; sign in, choose an organization, add billing if required, and generate a key under API keys. c-ai.chat is independent of Anthropic and the official Claude product. For broader setup context, start with our Claude API guide.

• The short answer
• How it works
• What it costs
• Limits and gotchas
• FAQ
• The honest take
• Sources

The short answer

To get a Claude API key, open platform.claude.com, sign in with an Anthropic account, create or select an organization, add a payment method if prompted, then generate a new key from the API keys area. Copy the key when it appears. You may not be able to view the full secret again.

The key is for Anthropic’s developer API. It is not the same as a Claude chat subscription. If you only want to use Claude in a browser, you do not need an API key. If you want your app, script, workflow, or backend to call Claude, you do.

export ANTHROPIC_API_KEY="sk-ant-..."
export ANTHROPIC_VERSION="set-this-from-the-current-docs"
export CLAUDE_MODEL="set-this-from-the-current-model-list"

curl https://api.anthropic.com/v1/messages 
  -H "x-api-key: $ANTHROPIC_API_KEY" 
  -H "anthropic-version: $ANTHROPIC_VERSION" 
  -H "content-type: application/json" 
  -d '{
    "model": "'"$CLAUDE_MODEL"'",
    "max_tokens": 200,
    "messages": [
      {"role": "user", "content": "Write a one-sentence product description."}
    ]
  }'

If the request succeeds, the API returns a message object with Claude’s generated text and token usage. If it fails, check the model ID, account status, billing setup, request headers, and rate limits before changing your application code.

How it works

A Claude API key identifies your organization when your software calls Anthropic’s API. Your app sends HTTPS requests to the Messages API with the key in the x-api-key header, a supported model ID, a message array, and settings such as max_tokens. Anthropic’s developer documentation is the source of truth for request formats, SDKs, model IDs, authentication headers, and versioning.

Treat the key like a password. Store it in a backend secret manager. Rotate it when staff leave or a repository leak is suspected. Use separate keys for local development, staging, and production. If your team is building more than a test script, assign ownership for billing, error monitoring, and model changes. Our overview of Claude features can help you decide which capabilities matter before you build.

1. Create or sign in to an Anthropic account

   Go to platform.claude.com. Use the account you want tied to your API organization, billing, and usage history.

2. Select the right organization

   If you belong to more than one organization, check the organization switcher before creating a key. Usage, limits, and invoices are scoped to the selected organization.

3. Add billing if required

   The console may require a payment method before API traffic is allowed. This is separate from any Claude chat plan.

4. Generate the key

   Create a new API key, give it a clear name such as prod-web-backend or local-dev, then copy the secret into a password manager or secret manager.

5. Test with a small request

   Call the Messages API with a short prompt and low max_tokens. Confirm that authentication, model access, and billing work before sending large documents or automated traffic.

6. Move the key out of code

   Use an environment variable such as ANTHROPIC_API_KEY. Never ship the key in frontend JavaScript, mobile app bundles, public repositories, or shared notebooks.

Product names and API IDs are not always identical. Use Anthropic’s model documentation for exact identifiers. Our Claude models guide explains how the main models compare at a practical level.

What it costs

Claude API access is usage-based. Anthropic charges separately for input tokens sent to the model and output tokens generated by the model. Claude chat plans do not include API usage. For plan-level context, see our independent Claude pricing guide, then confirm live API details in Anthropic’s official API pricing documentation.

Token pricing makes small tests cheap and uncontrolled automation expensive. A short classification request may cost very little. A workflow that sends long documents, tool results, chat history, and large outputs can grow quickly. Measure both sides: input and output.

Start with Claude Sonnet 4.6 unless you have a clear reason to use Opus 4.7 or Haiku 4.5. Sonnet is the usual default for assistants, drafting, analysis, coding help, and tool use. Use Opus for the hardest reasoning or quality-sensitive tasks. Use Haiku for high-volume background tasks where speed and cost matter most.

Limits and gotchas

Most Claude API key problems are not caused by the key creation screen. They come from billing state, organization selection, model IDs, rate limits, or unsafe key handling.

• Chat subscriptions do not replace API billing. Paying for a Claude chat plan does not automatically cover API usage. The developer platform has its own billing and usage controls.
• The organization matters. If you create a key under the wrong organization, requests may use the wrong billing account, limits, or model access.
• Rate limits can stop valid keys. A correct key can still receive rate-limit errors during bursts or high-volume jobs. Handle retryable errors with backoff.
• Model IDs must match the API. Use the identifiers listed in Anthropic’s model documentation. Do not assume a display name is the exact API string.
• Model availability can vary by account. Some models, betas, or higher-capacity features may require account access, platform settings, or plan changes.
• Region restrictions can block access. Anthropic services are not available in every country or region. If signup or billing is unavailable, consult Anthropic Support.
• Client-side exposure is a serious mistake. Anyone who sees your key can spend against your account until you revoke it. Never put the key in browser code, public repositories, mobile binaries, or shared demos.
• Long context is useful but not automatically cheap. 1M context can help with large files and repositories, but huge prompts increase cost and latency. Use retrieval, chunking, caching, and concise prompts where possible.
• Status incidents can look like app bugs. If requests fail across environments, check status.claude.com before rewriting code.

If you are using Claude in developer workflows, keep the API key separate from project files. Secrets belong in environment variables or managed secret stores, not in source control. For related setup material, see our Claude resources.

FAQ

If you are still deciding whether API access is the right route, compare it with the official Claude web app. The web app is simpler for personal use. The API is the right choice when you need automation, integration, logging, custom interfaces, or backend control. Our Claude FAQ covers more product-level questions.

The honest take

Getting a Claude API key is straightforward: use Anthropic’s platform console, create the key, store it securely, and test it with a small request. The important part is not the button you press. It is understanding that the key controls paid API usage and must be protected like any other production secret.

For most developers, the sensible path is to start with Claude Sonnet 4.6, set conservative token limits, log usage, and add prompt caching or the Batch API only when the workload pattern justifies it. Use c-ai.chat as an independent reference, but use Anthropic’s official docs for exact API syntax, model IDs, account controls, and live pricing.

Independent guide. Not affiliated with Anthropic. For the official Claude product, visit claude.ai.

Last updated: 2026-05-12

This article is part of the Claude API for developers hub on c-ai.chat.