How to Get Claude Free API Access: What's Real

Here's the straight answer on how to get Claude free API access, as of September 2026:

• New accounts get a small amount of free credits to test the API. That's Anthropic's wording on its pricing page, and it doesn't publish the amount.
• After that, you prepay. API usage draws down credits you buy in advance. When they run out, calls stop until you add more.
• Claude Pro and Max don't include API access. Anthropic's help center says so directly. A subscription does cover Claude Code, which is often what people actually wanted.
• Anthropic runs credit programs for AI safety researchers and for funded startups. Both require an application.

So the useful question isn't "where's the free tier." It's how to get a key, whether you qualify for a program, and how to keep the bill small.

How to get a Claude API key

The Claude Console now lives at platform.claude.com. The old console.anthropic.com address redirects there.

1. Sign up for a Console account. It's a separate product from the Claude app, with its own billing, even if you already pay for Claude.
2. Create a key under Settings → API keys → Create key. You name it and choose an expiration. Anthropic's authentication docs recommend identity-backed personal or service-account keys for new integrations.
3. Buy credits under Settings → Billing once the starting credits are gone. Auto-reload can top up when your balance drops below a threshold you set.

Three details worth knowing before you buy. Per Anthropic's billing article, purchased credits expire one year after purchase and are non-refundable. You're only charged for successful calls. And a request your client abandons partway through is still charged if it was on track to succeed.

Keep the key in an environment variable or a secrets manager, never in code you commit. Setting an expiration limits how long a leaked key stays usable. If you suspect a leak, you can disable the key (reversible) or delete it (permanent) from the same page.

Rate limits for new accounts

API limits come in usage tiers: Start, Build, Scale, and Custom. Each sets requests-per-minute and tokens-per-minute limits per model, and Start, Build, and Scale also carry a monthly spend cap. According to the rate limits documentation, new organizations may begin in an Evaluation tier with lower limits that rise automatically as you build usage history. If your first test script hits 429 errors quickly, that's a likely reason.

Official ways to get free Claude API credits

Beyond the starting credits, two Anthropic programs are worth knowing about. Neither is a free tier; both are applications with eligibility rules.

Enterprises evaluating Claude can also ask Anthropic's sales team about extended trials, per the pricing FAQ.

Claude Pro API access: what a subscription does cover

This is the most common mix-up. Anthropic's help center puts it plainly: a paid Claude subscription "doesn't include access to the Claude API or Console."

But the subscription covers more than chat. Pro and Max include Claude Code, in the terminal and in supported IDEs, sharing your plan's usage limits. If you wanted "API access" to write code with Claude, that's usually the better door. Our guide to installing Claude Code walks through setup.

One billing trap: if an ANTHROPIC_API_KEY environment variable is set on your machine, Claude Code uses that key instead of your subscription, and you pay API rates. Unset it if you meant to use the plan.

There's also a gray zone Anthropic is still working out. Its help article on the Agent SDK with a Claude plan says that, after a June 15 update, Agent SDK usage, claude -p, and third-party apps built on the Agent SDK still draw from your subscription's usage limits. Anthropic says it's revising that arrangement and will share changes before they take effect. None of that gives you an API key or Console credits, but it's why "can I use Pro programmatically" has a more complicated answer than it used to.

If you just want to try Claude without paying, the Free plan on claude.com is the intended route. It's a real plan with usage limits, not a workaround. If you're comparing no-cost options across assistants, our Claude alternatives guide covers what's genuinely free elsewhere.

Claude key for Janitor AI: what you're actually doing

Plenty of "free Claude API" searches come from people who want Claude inside Janitor AI, a third-party character-chat site. Janitor's own proxy setup guide has you add a configuration in its API settings, including a model name, a proxy URL, and an API key. Its route to Claude in that guide runs through OpenRouter, a separate service with its own credits and billing. In that setup Janitor isn't the one supplying Claude: you bring a key, and you pay whoever issued it.

What follows from that:

• Every token is billed to the key's owner. Anthropic's Commercial Terms, which govern API keys, make the customer responsible for all activity under its account. Chat frontends typically resend the conversation so far with each message, so a long chat costs more per reply as it grows.
• Give the app its own workspace. Every Console organization has a Default Workspace, and you can't set limits on it. Create a separate workspace for the third-party app and give it a monthly spend limit, so a runaway session stops at a number you chose.
• The Usage Policy still applies. Anthropic's Usage Policy covers anyone submitting inputs to its services, including through passthrough access. A third-party frontend doesn't change what's allowed on your key.
• "Free Claude keys" are someone else's account. A key posted on a forum or Discord bills its owner, who can disable or delete it from the Console at any moment. They tend to stop working without warning, and a site offering them in exchange for your own key is asking you to hand over your billing.

Making the bill small instead of zero

Anthropic documents several cost levers, and two of them stack. All figures below are from the pricing page as of September 11, 2026.

Prompt caching. If requests share a long, unchanging prefix, like a system prompt, a document, or a set of examples, a cache hit costs 10% of the base input price (2.5% on Fable 5.1). On the rate-limit side, cached reads don't count toward input-tokens-per-minute on current models, so caching raises your effective throughput too.

The Batch API. Work that doesn't need an answer right away can go through the Batch API at 50% off input and output.

Stack them. The pricing FAQ says Batch API and prompt caching discounts can be combined.

Pick the cheapest model that passes. Base prices run from $1 / $5 per million input/output tokens on Haiku 4.5 up to $10 / $50 on Fable 5.1. Anthropic's guidance on choosing a model suggests efficiency-first for cost-sensitive builds: start on Haiku 4.5, test properly, upgrade only where you find a real gap. Claude Haiku vs Sonnet shows where that gap usually appears, and which Claude model is best for coding covers the top of the range.

Lower effort before switching models. On models that support the effort parameter, Anthropic's guidance is that tuning effort is often a better lever than changing models.

Set a spend limit you'd be comfortable hitting. You can cap each non-default workspace, and the organization as a whole, below your tier's cap.

Next steps

If you're building something, create the Console account, use the starting credits to test, set a workspace spend limit, and turn on caching early. If you wanted Claude for coding, check whether a Pro or Max plan's Claude Code access covers you before buying API credits. And if a site is offering you a free Claude key, treat it as someone else's bill, not a loophole.

Getting a key takes minutes; what usually stalls is the step after, turning an empty key into a tool you use every week. With Taku, the starting point is an AI app or agent that already runs: you mirror it into your own desktop workspace and remix it, rather than beginning from a key and a blank file. Taku is in Beta, and the Mac app is available now.

FAQ

Is the Claude API free?

No. New accounts get a small amount of free credits to test it. After that, you buy prepaid credits, unless you qualify for one of Anthropic's credit programs.

Does Claude Pro include API access?

No. Anthropic's help center says a paid subscription doesn't include the Claude API or Console. Pro does include Claude Code, which shares your plan's usage limits.

Can I use a Claude API key with Janitor AI?

Janitor AI's proxy settings take an API key. Whatever it generates on an Anthropic key is billed to you, under Anthropic's terms. Put it in a separate workspace with a spend limit.

Where do I get a Claude API key?

In the Claude Console at platform.claude.com, under Settings → API keys.

Why am I hitting rate limits on a new account?

New organizations may start in an Evaluation tier with lower limits, which rise automatically as your usage history builds.