Starter
For one engineer tracking one stack.
$30/mo
$30 of usage included every month
Request access- 1 seat
- Local CLI + Throttle Cloud history
- Cost reports
- Email support
Early access · plans open by request
Start free with $10 of credits. No card.
Credits pay for Throttle Cloud. Every plan includes its price in usage, every month.
Pro
For teams that change configs often.
$50/mo
$50 of usage included every month
Request access- Up to 3 seats
- Everything in Starter
- Autotune. Throttle searches serving configs and keeps the cheapest that passes the output checks.
- CI cost gate. Fail a PR if $/M gets worse beyond noise.
- Alerts
Team
For an org running many models.
$100/mo
$100 of usage included every month
Request access- Unlimited seats
- Everything in Pro
- Shared org workspace
- Slack support
- Priority Cost Index data
Enterprise
For large GPU fleets.
Custom
Talk to us- Volume pricing
- A hands-on pilot. We measure and fix your stack, and you pay a share of verified savings. How it works
- Invoicing
- Dedicated support
The Throttle CLI stays free and open source (MIT). Credits pay for Throttle Cloud: history, Autotune, CI gates and team features.
Usage
Simple rates.
Plan credits are spent at these rates. Beyond them, overage is billed at the same rates.
| What | Price |
|---|---|
| Cloud checkOne measured check, stored in Throttle Cloud | $0.25 |
| Autotune trialOne serving config tried and verified | $1.00 |
| CI gate runOne cost check on a pull request | $0.10 |
Top up credits any time. 1 credit = $1 of usage.
FAQ
Questions.
What is a credit?
One credit is one dollar of usage. You spend credits on Cloud checks, Autotune trials and CI gate runs at the rates above.
What happens when credits run out?
Cloud features pause until you top up or upgrade. The CLI keeps working, locally, for free.
Can I cancel?
Any time.
Do you see my prompts?
No. Checks run on your side, against your own endpoint. Throttle Cloud stores the numbers, not your prompts or responses.
Not sure which plan?
Grab 20 minutes. We’ll look at your stack and tell you.