Anthropic announced on October 7, 2026 that Claude Max and Team subscriptions now include monthly API credits. The Claude Platform release notes establish the change for subscribers building their own applications and agents.
For an existing subscriber, the benefit can fund a prototype’s eligible usage. Whether it covers the whole project depends on the workload, remaining balance and account limits—not just the allowance advertised for the plan.
Who qualifies and how much you get
Anthropic’s eligibility rules require an active, good-standing subscription. New subscribers wait seven days. Mobile purchases and discounted Nonprofit/Scientists Team plans qualify; Free, Pro and Enterprise do not. Rollout takes several days.
Subscription | Monthly API credit, USD |
|---|---|
Max 5x | $100 |
Max 20x | $200 |
Team (shared pool) | $20 per Standard seat + $100 per Premium seat; $500 combined ceiling |
Credit amounts, not subscription prices. Source: Anthropic’s plan table.
Team grants use seats present at the billing month’s start, including the first grant; added seats count next month. Seat-count rules.
Illustrative calculation: monthly Team credit = min($20 × Standard seats + $100 × Premium seats, $500). Three Standard seats plus two Premium seats yield $260; six Premium seats still yield $500, not $600. Once the ceiling is reached, another seat adds no API allowance. Buy it for the app access your team needs, not for a larger API pool.
Claim into the right Console organization
Each plan links to one Console organization, and each organization accepts credits from one plan. Changing the link requires support. Linking restrictions.
On claude.ai, open Settings → Billing for Max, or Organization settings → Billing for Team.
Claim as the Max subscriber or a Team Owner/Primary Owner, with Console Owner, Admin or Billing access.
Select Link organization, choose or create the destination, review the credit terms and confirm. No Platform payment card is required. Claiming instructions.
Check the issued amount and expiry in Console Billing before budgeting. Grants follow billing cycles, monthly on annual plans; unused credit expires without rollover. Credit timing.
Do not schedule a full month of work from the day you click Claim without checking that expiry. The documentation does not fully resolve late-first-claim timing.
What the credits pay for
Covered: Claude API, Managed Agents, Agent SDK and Playground. Excluded: Claude Code, extra usage in Claude apps, Claude Platform on AWS, Bedrock, Google Cloud and Microsoft Foundry. Existing app limits are unchanged. Coverage rules.
For SDK projects, use an API key from the credited organization. Anthropic’s October 7 SDK notice says the earlier June SDK credit is no longer available; this Console-linked benefit replaces it. Do not treat an old SDK-credit tutorial as the current claiming procedure.
A worked prototype budget—not a benchmark
Anthropic lists standard global Sonnet 5.5 rates of $2 per million input tokens and $10 per million output tokens. Current API pricing.
Illustrative assumptions: each request bills exactly 10,000 uncached input tokens and 1,000 output tokens, including any billed thinking. Exclude tools, cache operations, hosting, retries, taxes, other users and additional feature charges. No claim is made that a real task fits those counts.
Token cost per request = (10,000 × $2 + 1,000 × $10) ÷ 1,000,000 = $0.03.
Available eligible budget | Whole requests at $0.03 each |
|---|---|
$100 | 3,333 |
$200 | 6,666 |
$500 | 16,666 |
These are calculated token-budget ceilings, not successful tasks or guaranteed throughput. An agent session can make many requests. Growing context, additional attempts and tool charges change the total; external infrastructure needs its own budget.
For asynchronous work, published batch rates of $1 input/$5 output per million tokens make the same assumed request $0.015. That option helps only if waiting for batch results fits the application. Batch pricing.
Before a pilot, define what counts as an accepted result. Track spend across failed attempts as well as accepted ones; divide total cost by accepted outcomes only when at least one passes. This proposed evaluation is more useful than treating every funded request as completed work.
Keep the balance, spend cap and rate limit separate
Included credits are consumed first. Auto-reload monitors purchased credit, so it can charge while included credit remains. Review its settings. Application order.
Credit-funded usage counts toward calendar-month spend caps, whose reset may differ from credit renewal. Linking grants at least Start tier; allowances above $200 grant at least Build. Tier rules.
Start’s monthly spend cap is $500 and Build’s is $1,000. These are ceilings, not extra credit. A tier-cap pause lasts until 00:00 UTC on the next month’s first day unless the limit is raised. Short-period request and token throttling are separate. Spend and rate limits.
When included credits run out, purchased balances can continue funding requests; otherwise requests stop. Sales-invoiced overage is billed normally, not to the subscription. Exhaustion rules.
For a pilot, have an administrator create a non-default workspace and set a project limit. Workspaces retain centralized billing, not separate credit grants. The Default Workspace cannot have its own limits, and organization-wide ceilings still apply. Workspace controls.
When a job stops, inspect the balance and the actual limit error before adding funds. More purchased credit does not remove a spend cap or make burst traffic acceptable. Keep application state resumable so a budget interruption does not require restarting the whole task.
Claiming a benefit is different from buying a plan
The web price of Max 5x is $100 monthly; Max 20x is $200. Mobile prices can differ. Max subscription pricing. Credits are non-refundable and have no cash value. Credit terms.
Consider a full recurring Max 5x month after eligibility, with $30 of otherwise payable eligible API usage. Ignore taxes, discounts and any value assigned to the app. An existing subscriber offsets $30 of additional API spending—not $100. Someone buying the plan solely for that workload pays $100 to use $30 of the benefit, leaving $70 unused.
That arithmetic argues for claiming an existing entitlement before upgrading, and comparing direct API purchasing for intermittent, low-volume projects. Value the Claude app separately; the nominal credit alone does not prove that a new subscription is the right purchase.
Choose one bounded task before allocating the balance. RohitAI’s chatbot, workflow or agent decision guide helps match the design to the task and define acceptance checks. Then budget that workload, rather than expanding it merely because credits are available.
Methodology: This AI-assisted guide uses Anthropic’s published release notes, support pages, pricing and administration docs, checked October 7, 2026. All examples are arithmetic or proposed evaluation steps. No authenticated claim, API workload, billing transition or hands-on benchmark was performed.