Every plan reaches the same models through the same gateway. The difference is how many keys you run and how much credit comes with the month.
Rates sit above each publisher's list price. That spread is what you are buying: a gateway that forgets you. See the rates →
Top up, spend, stop whenever.
- List rates +30%Per token, charged per request
- 1 keyRotate any time
- Secret masking onRequested on supported text endpoints
- Pay USDC on BaseApprove payments through your Privy wallet
For one person running real work through agents.
- $30 of credit each monthCredit lasts for the billing period
- 10% off every rateList +17% instead of +30%
- 25 keys with budgetsModels, endpoints and monthly ceilings
- Request log and CSV exportCosts and token counts, never content
For teams that need more keys and control over text routing.
- $150 of credit each monthCredit lasts for the billing period
- 15% off every rateList +10.5%
- 200 keys, scopedPer model, per endpoint, per budget
- Provider selection for textChoose which providers a key may use
- Request log and CSV exportCosts and token counts, never content
Plans include credit, not a quota. Heavy use spends the credit faster, at the plan's rate, then continues from your top-ups.
Model rates
Per million tokens, pay-as-you-go. Pro and Sealed pay 10% and 15% less. The catalog is wider than this table; anything /v1/models returns is available at the same markup. Refreshed daily.
| Model | Publisher | Context | Input | Output |
|---|---|---|---|---|
| Claude Opus 4.8claude-opus-4.8 | Anthropic | 1000k | $7.80 | $39 |
| Claude Sonnet 4.6claude-sonnet-4.6 | Anthropic | 1000k | $4.68 | $23.40 |
| Claude Haiku 4.5claude-haiku-4.5 | Anthropic | 200k | $1.30 | $6.50 |
| GPT-5.5gpt-5.5 | OpenAI | 1000k | $6.50 | $39 |
| GPT-4o Minigpt-4o-mini | OpenAI | 128k | $0.20 | $0.78 |
| Gemini 3.1 Progemini-3.1-pro | 1049k | $2.60 | $15.60 | |
| DeepSeek V4 Prodeepseek-v4-pro | DeepSeek | 1000k | $0.27 | $0.54 |
| GLM 4.7glm-4.7 | Z.ai | 128k | $0.78 | $2.86 |
| MiniMax M2.1minimax-m2.1 | MiniMax | 198k | $0.39 | $1.56 |
| OpenAI GPT OSS 120Bopenai-gpt-oss-120b | OpenAI | 128k | $0.09 | $0.36 |
How billing works
Charged per request
Each response carries its cost in a header. Streamed responses carry token counts in the final usage event and the cost in your request log.
Credit first, then top-ups
Plan credit is spent before anything you added. Top-ups never expire.
Monthly key ceilings
New requests stop with a 402 once spend and pending reservations reach the ceiling. A final request can exceed the remaining amount.
USDC settles at finality
Sign in and pay USDC on Base through your Privy wallet. Your balance updates after onchain confirmation.
No seats
Use separate keys for agents, jobs or teammates, within your plan’s key limit.
Cancel by stopping
Drop the plan and pay-as-you-go keeps working on whatever credit remains.
Questions
Why is this more than going direct?
Because direct means an account in your name, prompts attached to it, and a vendor that knows your users. You are paying for the gap between you and them.
What do you keep?
Key id, model, token counts, cost, timestamp. Nothing you typed and nothing the model said. The full list is in the docs.
Which models?
The ones above plus everything the catalog returns from /v1/models. When a provider adds a model it appears without a release note from us.
Can I use Claude Code?
Yes. Three environment variables, documented here. Cursor, Aider, Continue and OpenCode work the same way.
Why USDC only?
A card means a processor, a name and a billing address on file. A stablecoin transfer to an address derived for you alone means none of that.
What can the model provider see?
The request body arrives under our gateway credentials. We do not forward your Shadr key or client IP headers. Text requests ask for secret masking, but personal information you put in a prompt can still reach the provider.
Is there a free tier?
No free tokens. Pay-as-you-go has no monthly fee; send as much or as little USDC as you like.