Stop hitting usage limits.
Pay per token, not per plan.
One OpenAI- and Anthropic-compatible API. Prepaid, billed per token, with a receipt for every request and live prices next to the official list price.
No card required · The $1 free credit lasts 7 days · Credits you buy never expire
curl https://newtons.dev/api/v1/chat/completions \
-H "Authorization: Bearer $NEWTONS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5-5",
"messages": [{"role": "user", "content": "Hello!"}]
}'
Built for people who run out of plan before they run out of work
Coding agents, scheduled jobs and daily heavy builders burn through weekly and five-hour caps. Newton's bills the tokens you use, so work keeps going until your balance says stop.
Flat-rate AI plans
$20–$200/mo- Pay the same whether you use it or not
- Five-hour and weekly caps stop agents mid-task
- One vendor's models only
- No cost per request visibility
Newton's
Pay per token- No weekly cap: you pay for tokens used
- Published price for every model
- 9 models, one balance
- Exact per-request receipts
Prices per million tokens
Newton's price next to the official list price. Cached-input rates are on each model page.
Every request shows its work
No mystery monthly bill. Each request is priced from the published rate, charged exactly, and saved in your console.
- Input
- 10,000 tokens × $0.20 / 1M
- $0.0020
- Output
- 2,000 tokens × $1.00 / 1M
- $0.0020
- Request fee
- flat
- $0.0001
- Total charged
- $0.0041
- Balance after
- from your $1 starter credit
- $0.9959
An example at today's published Sonnet 5.5 rates. Your console shows the real receipt for every request you run.
Running in 3 steps
Get a key
One click, no card. Test with the $1 starter credit, then add a balance when you're ready.
Change your base URL
Change one line in your OpenAI SDK. Your existing code keeps working.
Watch every request
Each accepted request gets a receipt with its exact cost. When the balance reaches zero, requests stop instead of surprising you.
Use the tools you already run
Point any OpenAI- or Anthropic-compatible tool or SDK at Newton's by changing its base URL and key.
Common questions
Why is Newton’s priced below the model makers’ list prices?
Newton’s buys capacity from upstream API marketplaces and capacity resellers that sell below the model makers’ list prices, then adds a margin. Wherever we have verified the official list price, we show it next to ours with a source and a date, so you can check the gap yourself. Where we have not verified it, we show no comparison.
Does Newton’s store my prompts?
Newton’s keeps the request records needed for billing and receipts, such as model, token counts, cost and timing. Your prompts are forwarded to an upstream supplier and the model provider to produce a response, and those parties apply their own retention policies. We do not offer a zero-data-retention guarantee, so do not send data you are not allowed to share with third-party processors.
How much latency does Newton’s add?
A gateway adds a network hop, and requests pass through an upstream supplier, so expect more latency than calling a model provider directly. Time-to-first-token varies by model and route, and we do not publish a latency guarantee.
What happens if an upstream route fails or is rate limited?
Newton’s does not guarantee uptime. If an upstream route returns an error, your client receives that error and should retry with backoff. Each accepted request gets a receipt with its exact cost, and your prepaid balance is a hard spending limit.
Run your agents without the caps
Prepaid balance. A receipt for every request. Pay only for tokens you use.
Get an API key · $1 free credit →