Using Cloudly
Spend limit
Set a monthly limit on model spend in Settings → Spend limit. It covers model usage on your API keys only; your VM is billed by your cloud provider, and turns on a Claude or ChatGPT plan don't count.
What happens as you spend
| Spent | What happens |
|---|---|
| Under 80% | Nothing interrupts you. Each turn's cost is recorded when the agent reports it. |
| 80% | A banner in the app and one email for the month. Turns keep running. |
| 100% | New turns are refused until you raise the limit or the month ends. |
A turn that is already running when you reach 100% finishes, so a task isn't left half done. Claude Code turns stop themselves at 25% over the limit and keep the work they finished on the branch. Codex and Gemini can go over by that one turn.
Months follow UTC, and both alerts re-arm when you save a new limit.
Exact and estimated costs
- Claude Code reports the exact cost of every turn.
- Codex and Gemini CLI report tokens. Cloudly prices them from a table you control and marks the result as an estimate, shown with a
~.
Set your own prices, in US dollars per million tokens, in the configuration:
PRICE_CODEX_INPUT_PER_MTOK=1.25
PRICE_CODEX_OUTPUT_PER_MTOK=10
PRICE_GEMINI_INPUT_PER_MTOK=0.30
PRICE_GEMINI_OUTPUT_PER_MTOK=2.50
Email alerts
Alerts are in-app by default. To also get email, add a Resend API key under Settings → Keys and connections and an alert address under Settings → Spend limit. Set the sender with ALERT_FROM in the configuration. SMTP isn't supported yet.