Claude Max API credits: $100 or $200 free every month, claimed once

In short

• It's free. The credit is included in your Claude Max plan, at no extra cost and with no card to add.
• It's added to your existing account. No new subscription: the credit goes to an API space, the Claude Console, that you link once to your Claude account.
• It's every month. $100 on Max 5x, $200 on Max 20x, paid at each plan renewal, monthly on annual plans too.
• It doesn't roll over. Whatever you don't spend disappears at the end of the month.
• It doesn't arrive on its own. Until you make the link, no credit lands anywhere.

This credit does not give you more Claude or more Claude Code. It is for calling Anthropic's models from your own scripts, tools and agents.

Claim it in three steps

1. On claude.ai, open Settings > Billing. In a web browser: Anthropic asks you to claim on the web, even if you subscribed from your phone.
2. In the "API credits" section, click "Link organization". Pick a Claude Console organization, or let the flow create one if you never had one. Accept the terms, and you're done.
3. Check that the credit arrived. On platform.claude.com, under Settings > Billing, the amount shows up in the "Promotional credits" section, with its expiry date.

Three things to know before you click:

• The link is permanent. A plan links to a single organization, and you can't change it yourself: only support can. If you already have a Console organization that belongs to a client or your employer, don't send your personal credit there. Create your own.
• You need seven days on the plan. A brand-new Max subscriber waits a week before the section shows up.
• The rollout is gradual. Anthropic started it the week of October 7. If you don't see the section yet, give it a few days before contacting support.

Who gets it

Pro is left out, unlike September's cloud credit.

What the credit pays for

Every model is included, from Haiku 5.5 to Opus 5.5. And your plan limits don't change: your chats and Claude Code sessions still draw on your Max allowance, as before.

Hermes, OpenClaw and other harnesses

Yes, they count. A harness such as Hermes Agent, OpenClaw or OpenCode, plugged into an API key from the linked organization, is pay-per-token API usage: exactly what the credit pays for. It's even the cleanest route, while signing in with a subscriber account has changed a lot since April.

How to plug them in, which model to give which agent, what it really costs over a month: we're working on a full article on the subject, coming soon.

Stay safe: never pay without meaning to

The credit is free, but the Console is a paid service. A few settings keep it free.

• Don't add a credit card. You don't need one to use the credit. With no payment method, nothing can be charged: once the credit runs out, calls stop until the next payment, and nothing is ever charged to your Claude plan.
• If a card is already on file, turn off auto-reload. In the Console, under Settings > Billing, switch off "auto-reload". Otherwise paid credits get bought as soon as your purchased balance drops below your threshold, even while included credit remains: auto-reload only watches the purchased balance.
• Cap the organization's spend. Still under Settings > Billing, in the "Spend limits" section, click "Adjust limit" and set a limit equal to your credit, $100 or $200. Spend paid by the credit counts toward that cap, so it stops everything beyond it. One catch: the cap resets on the 1st of the month, the credit on your plan's renewal date.
• One workspace per project, each with its own limit. Every key draws from the same balance. A limit per workspace stops a looping script from eating the whole month.
• Keep your keys secret. An API key gives access to your balance. Never paste it into a file pushed to GitHub or into a chat. If in doubt, delete it in the Console under API keys and create a new one.
• Glance at the Cost page during the first week, until you know what your uses cost.

If you cancel Max, new credits stop, but credit already received stays usable until it expires. Upgrading from Max 5x to Max 20x gives you a prorated top-up right away, then $200 a month.

What $100 buys

Working assumption: an everyday request with 2,000 input tokens and 500 output tokens, roughly one page sent and half a page back.

Haiku 5.5's price applies to prompts under 100,000 tokens; above that, it is five times higher. The Batches API, for anything that can wait a few hours, halves the bill. Prompt caching for repeated context does the rest. Track the spend on the Console's Cost page.

Five uses worth the credit

Your Max plan already covers chat and coding. The API credit is for what the plan doesn't do: run without you.

• Batch processing. Sort 3,000 support tickets, summarize a year of meeting notes, pull fields out of a stack of invoices. With Haiku 5.5 through Batches, a few dollars cover thousands of documents.
• Everyday scripts. A claude -p wired to a scheduled job that reads the error logs every morning or sends a news digest every Monday.
• An agent that runs with your laptop closed. Claude Managed Agents hosts the agent at Anthropic. Runtime costs 8 cents an hour on top of tokens, and the credit covers both.
• A personal agent. A harness like Hermes, plugged into your messaging app, working for you around the clock. See above.
• Your own benchmark. Compare Haiku, Sonnet and Opus on your task, with your data, before choosing. It's the cheapest way to know instead of guess.

Your first call

Without code, the Playground is enough: pick a model, write an instruction, read the answer and what it cost.

To go further, create an API key in the Console, inside a capped workspace as described above. A first call from the terminal:

export ANTHROPIC_API_KEY="sk-ant-..."

curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-haiku-5-5",
    "max_tokens": 512,
    "messages": [{"role": "user", "content": "Summarize these meeting notes in three points: ..."}]
  }'

One trap. Don't leave ANTHROPIC_API_KEY in your shell profile (.bashrc, .zshrc). Claude Code can pick it up instead of your subscriber login: your sessions would then be billed to the organization, which the credit doesn't cover for that use, and Claude Code would stop with "Credit balance too low". Pass the key to the command that needs it instead:

ANTHROPIC_API_KEY="sk-ant-..." claude -p "List the functions without tests in src/"

Sources

• Claude Haiku 5.5 announcement, Anthropic, October 7, 2026: amounts per plan, rollout this week, all models.
• "API credits for Max and Team plans", Claude Platform documentation, accessed October 8, 2026: payment every cycle, no rollover, steps, covered products, permanent link.
• "Monthly API credits for Max and Team plans", Claude Help Center, accessed October 8, 2026: seven days on the plan, mobile subscriptions, interactive Claude Code excluded.
• "Use the Claude Agent SDK with your Claude plan", Claude Help Center, October 7, 2026 update: Agent SDK and claude -p coverage.
• Claude Platform pricing, accessed October 8, 2026: per-model prices, Batches discount, Managed Agents.
• Third-party harness restrictions on subscriptions, April and May 2026: reported by Let's Data Science and MindStudio, among others.
• Rate limits and spend limits, accessed October 8, 2026: setting the spend limit, reset on the 1st of the month, workspace limits.