How to Get an Anthropic API Key

If you want to use Claude through a bring-your-own-key client, you need Anthropic API access, not only a Claude consumer subscription.

Anthropic operates its API through the Anthropic Console. The Console is where API users manage keys, billing, usage, and developer access.

This distinction matters because Claude’s consumer plans and Anthropic’s API are separate products with separate billing.

1. Open the Anthropic Console

Go to the official Anthropic Console:

console.anthropic.com

Sign in or create the account/organization you want to use for API access.

If you already pay for Claude on the consumer website, do not assume that subscription automatically funds API requests. Anthropic documents the consumer product and API Console as separate services.

2. Make sure the account can use the API

Before a new key is useful, the organization needs API access and billing set up according to Anthropic’s current Console requirements.

Depending on your account and organization, this can include adding billing information or usage credits.

Anthropic changes product and billing details over time, so use the Console as the source of truth rather than following an old screenshot-based tutorial.

Official background:

• How can I access the Anthropic API?
• I have created an account in Console and I want to start using the API. What should I do?

3. Create an API key

In the Console, open the API keys area and create a new key for the account or workspace you intend to use.

Anthropic’s current support documentation refers to a Create Key action on the API Keys page.

Give the key a name that helps you understand where it is used. For example:

BYOKchat

or:

Personal AI client

A descriptive name makes it easier to revoke one client later without wondering which applications depend on the key.

4. Copy the key once and treat it as a secret

An API key is not a harmless identifier. It authorizes API usage on your Anthropic account and can create charges.

Treat it with the same care you would give another sensitive credential:

• do not paste it into public issues or chat messages;
• do not commit it to Git;
• do not include it in screenshots;
• do not store it in a plaintext notes file;
• do not send it by email for support;
• do not reuse a leaked key after deleting the public copy.

Anthropic’s own key-security guidance explicitly warns that exposed keys can be used by someone else and can incur charges on your account.

See API Key Best Practices: Keeping Your Keys Safe and Secure.

5. Add Anthropic as a provider in your BYOK client

A BYOK client will usually ask for at least:

• the provider type;
• the API key;
• a model choice.

For a native Anthropic integration, the client talks to Anthropic’s Messages API rather than pretending Claude is an OpenAI-compatible endpoint.

That distinction can matter for features such as:

• Anthropic-specific streaming behavior;
• reasoning/thinking support where available;
• tool calling;
• provider-specific continuation metadata;
• model discovery or curated model selection.

In BYOKchat, Anthropic is a first-class provider across the current web and native product surfaces. Credential storage depends on the platform: native iOS/macOS can use Apple Keychain, while the web app keeps the user-supplied credential inside the browser-origin local storage boundary.

For that security difference, read Where Does a BYOK Web App Store Your API Key? and BYOK Web App vs Native App: What Changes?.

6. Test the connection before depending on it

After saving the connection, run a simple test before starting important work.

If authentication fails, check the basics first:

• Did you copy the complete key?
• Was the key deleted or rotated in the Console?
• Does the account have API billing/credits available?
• Does your organization role allow the required API access?
• Is the selected model currently available to the account?

If the provider returns a quota, billing, or rate-limit error, the Anthropic Console is the right place to inspect account-level status because a BYOK client only sees the error returned by the API.

Claude subscription vs Anthropic API

This is the most common source of confusion.

A paid Claude plan gives you benefits in Anthropic’s consumer Claude product. API usage through Anthropic Console is billed and managed separately.

So this workflow:

Claude subscription → copy a hidden API key → use it elsewhere

is not how the API works.

The correct mental model is:

Anthropic Console account
  → API billing/access
  → API key
  → BYOK client
  → Anthropic API

For the current subscription/API distinction, see Does Claude Pro Include Anthropic API Access?.

Should you create one key per app?

Using separate keys for separate applications can make credential management easier.

If you later stop using one client, you can revoke its key without rotating every integration that uses your Anthropic account.

It can also make usage audits easier when the provider reports activity by API key.

For personal use, you do not need to create a complicated key-management system, but a descriptive dedicated key is a useful habit.

What to do if the key is exposed

If you suspect the key has leaked, revoke/delete it in Anthropic Console and create a replacement.

Do not rely on deleting the message, file, or Git commit where the key appeared. Once a credential has been exposed, you should assume it may have been copied.

Then:

1. create a new key;
2. update the client that needs it;
3. review Anthropic usage for unexpected activity;
4. keep the replacement out of version control and public communication.

You do not need to share the key with the website or support team

A BYOK product marketing site or support workflow does not need your provider credential.

The key belongs only in the client surface you deliberately choose to make authenticated Anthropic requests. On the web, that means understanding the browser-origin trust boundary; on native Apple platforms, that can mean Keychain-backed secret storage.

If a support workflow asks you to send your raw API key to diagnose a connection problem, do not do it. Error messages, app version, provider configuration without secrets, and sanitized diagnostics should be enough to investigate most problems.

Keep the provider relationship yours

The main reason to use your own Anthropic API key is not simply to unlock Claude in another interface.

It means:

• the Anthropic account is yours;
• the API usage is billed to your provider account;
• you can revoke the credential yourself;
• you can inspect provider-side usage;
• you can move between compatible clients without asking one chat service to migrate your provider relationship.

That is the practical meaning of BYOK: the client is your workspace, while the provider account remains under your control.