Anthropic Startup Program: credits and API budgeting
The Anthropic Startup Program is accepting applications, but its earlier Team and $1,000 API offers are over capacity while Anthropic reshapes the program. Budget your Claude API usage at normal rates until a usable credit balance appears in your company's account.
Program availability can change. This independent guide explains the public terms and token-cost arithmetic; it does not promise acceptance or credits.
Who can apply, and what does VC backing change?
The official program page welcomes bootstrapped, pre-seed and venture-backed companies. Sign in to Claude Console and use the application linked there. You do not need a partner VC to use that public route. Anthropic aims to review applications within a week; that is a target, not a guaranteed decision date.
Selection remains discretionary. The official terms describe factors such as business traction, funding and Claude integration or usage, as well as geographic and legal restrictions. Check those terms against your company before applying. Funding is one consideration, not a universal requirement.
Startups backed by a partner VC may qualify for up to $100,000 in additional API credits through that VC. Ask your investor about its channel and your company's eligibility. That headline is a conditional ceiling; it is not the public application's standard award. Benefits are assigned to the company account.
Separate the three kinds of benefit
- Claude API credits: only an awarded, claimed and usable balance can offset eligible API usage.
- Claude subscriptions: Team access and other plan benefits have their own terms. A subscription is not a general API wallet.
- Startup Stack: Anthropic advertises up to $45,000 in discounts and credits from independent third-party companies. These are partner offers, not $45,000 of Claude API credit or cash.
Existing members can still access Startup Stack offers and program events. Anthropic also mentions monthly API credits for eligible Max and Team plans; check that separate plan entitlement in Console instead of adding it to an assumed startup grant.
Normal Claude API costs before any credit
Direct API Standard prices in USD per million tokens, checked against Anthropic's pricing documentation. Both models keep these rates across their full 1M context windows.
| Model | Uncached input | Output | Cache reads | 5-minute cache writes |
|---|---|---|---|---|
| Claude Sonnet 5.5 | $2 | $10 | $0.1 | $2.5 |
| Claude Opus 5.5 | $4 | $20 | $0.2 | $5 |
Eligible Batch work halves Standard input and output prices. Cache writes and hits have different rates, and a hit is not guaranteed. US-only inference adds a 1.1× multiplier across token categories. Tool charges, taxes, retries and billable thinking can change your total.
A monthly budget you can reproduce
Assume 10,000 requests a month, each with 10,000 uncached input tokens and 2,000 billable output tokens. Use Standard global processing, with no paid tools or retries. Sonnet 5.5 costs $0.02 for input plus $0.02 for output per request; Opus 5.5 costs $0.04 plus $0.04.
| Model | One request | 10,000 requests | Requests covered by a hypothetical $1,000 usable credit |
|---|---|---|---|
| Claude Sonnet 5.5 | $0.04 | $400 | 25,000 |
| Claude Opus 5.5 | $0.08 | $800 | 12,500 |
The last column illustrates arithmetic only. It assumes your account actually has $1,000 available for this usage and that you spend it before its expiry. It does not describe a currently guaranteed offer.
Open these exact models and inputs in the calculator, then replace them with your expected traffic. It estimates gross token costs before credits. Its cache toggle treats all input as existing cache reads; it does not model partial hits or the first cache write.
Turn an account balance into a spending plan
Record the amount awarded, the remaining usable balance, its expiry and which charges it covers. Check those details in your award and Console. The public program page does not give every applicant a universal credit amount or expiry period.
For eligible usage, estimate cash charges as the usage cost minus credits actually applied, with a floor of zero. A $400 token bill with $300 of applicable credit leaves $100 before other charges. A pending application leaves the full $400 in your budget. Once credit expires or runs out, the same workload still costs $400 at these rates.
Startup credits cover the first-party Claude API through Claude Console. They do not cover Amazon Bedrock, Google Vertex AI or another reseller's invoice. Members who receive API credits also receive higher rate limits, but that does not mean unlimited throughput. Check your account's limits before planning a launch.
What if I was approved but cannot see the offer?
A community discussion about the withdrawn offers shows why this distinction matters. User reports describe individual experiences; the current official FAQ is the source for program policy.
Anthropic says already claimed offers remain usable. It is re-reviewing applications, which may change application status. An approval email alone therefore should not become a credit line in your forecast. Check what was actually claimed in your account, and use the official program contact if the account and your award disagree.
AI API Prices editors checked the linked official pages on October 11, 2026. For a launch budget, start with the full Anthropic price table or the Sonnet 5.5 cost guide, then measure total usage per accepted result, including failed attempts.