On 7 October 2026, Anthropic started including monthly API credits in its Claude Max and Team plans. If you pay for Max or Ultra, you get $100 or $200 of credit every month to run your own apps and agents on the Claude Platform. Team plans get up to $500, shared across the team. The credits don't roll over, so every month you don't claim and use them is money left on the table. Most subscribers we talk to haven't noticed, and claiming takes about two minutes.
How much you get
The amount depends on your plan. Max includes $100 a month and Ultra includes $200. Ultra was called Max 20x until Anthropic renamed it, and Max was called Max 5x. On Team plans, each Standard seat adds $20 and each Premium seat adds $100, pooled across the team and capped at $500 a month. A team with three Standard and two Premium seats, for example, gets $260. Free, Pro and Enterprise plans aren't eligible.
From Anthropic's help centre, October 2026. Team credits are $20 per Standard seat and $100 per Premium seat, up to $500.
To put that in context, Haiku 5.5 costs $0.10 per million input tokens. One hundred dollars of credit is enough to have a fast model read hundreds of millions of words, or to run a capable agent on Opus 5.5 for a meaningful amount of real work every month.
How to claim them
You need to have been on an eligible plan for seven days. If you subscribed through the iPhone or Android app, claim in a web browser at claude.ai.
On Max, go to Settings, then Billing. On Team, the Owner or Primary Owner goes to Organization settings, then Billing. In the API credits section, select Link organization, choose an existing Claude Console organization or create a new one, review the credit terms, and confirm. You need the Owner, Admin or Billing role in that Console organization. You don't need to add a payment method on the Claude Platform. The credits then appear in the organization's balance, which you can check with the expiry date on the Claude Platform billing page.
Choose the organization carefully, because each plan links to one Console organization and you can't change it yourself afterwards. Changing it means contacting support.
What the credits cover, and what they don't
The credits work for the Claude API, including batch requests, the Console Playground, Claude Managed Agents and the Claude Agent SDK, with any available model. They are spent before any credits you buy. If you run out and haven't bought extra credits or turned on auto-reload, API requests simply stop until next month. Anthropic says the usage is never charged to your Claude plan.
The most important limit is that the credits don't pay for Claude Code when you use it interactively, whether in the terminal, an IDE, the desktop app or the web. They also don't cover extra usage in Claude or Cowork, or Claude through Amazon Bedrock, Google Vertex AI or Microsoft Foundry, and they don't change your plan's usage limits. Scripts that call Claude with your own API key from the linked organization are covered, including the Agent SDK and headless Claude Code runs started with claude -p.
Credits refresh with each billing cycle and expire at the end of it. If you cancel or downgrade, new credits stop, but the ones you have stay usable until they expire.
Three things worth building with them
You don't need to be a developer to get value here, especially if Claude writes the code for you. Start with a small automation that saves time every week. A feedback sorter is a good first project: a short script that sends every new survey answer or support ticket to Haiku 5.5, tags it by theme and sentiment, and writes the results to a spreadsheet. At Haiku's prices, it costs almost nothing to run.
Second, try a scheduled research agent. With the Agent SDK, you can build an agent that checks a list of competitor pages or industry news sources every morning and sends you a short summary. Ask Claude to write it with you step by step, then run it from your own computer or a cheap server.
Third, if your team keeps asking the same questions, build an internal helper that answers from your own documents. On a Team plan the pooled credit can cover this, and an owner can set workspace spend limits in the Console so one experiment doesn't use up everyone's credit.
What we would do this week
Claim the credits today, even if you have no plan for them yet, so they are ready when you do. Then pick one repetitive task that involves reading or sorting text and build the smallest possible version of an automation for it. If you aren't sure which model to use, our guide to choosing between Fable, Opus, Sonnet and Haiku will help, and our Kits page collects ready-made automation ideas by department.