Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.kodus.io/llms.txt

Use this file to discover all available pages before exploring further.

The monthly spend limit watches how much you’re spending on your BYOK models and notifies your organization’s Owners as that spend approaches a cap you set. It’s an early-warning signal, not a hard stop.
This is a notification, not a cap. Kodus never blocks or pauses reviews when the limit is hit. Spend keeps accruing on your provider account past 100%. To actually stop spend, set a hard limit in your model provider’s dashboard (OpenAI, Anthropic, Google, etc.) — that’s the only thing that can enforce it. Treat the Kodus limit as a heads-up so the provider cap is never a surprise.

​
What it tracks

  • BYOK models only. Spend is measured against the models you configured under Bring Your Own Key — the keys you pay your provider for directly. Models Kodus runs on its own side don’t count.
  • Priced at current rates, per model. Each model’s tokens are priced at that model’s current rate. There’s no snapshot — if a model’s price changes (catalog update or a manual edit), that model’s month-to-date usage is recomputed at the new rate. Switching to a different model doesn’t re-price the model you were using before.
  • Per calendar month (UTC). The window runs from the 1st of the month to the 1st of the next. Usage resets to $0 automatically at the start of each month — there’s nothing to reset by hand.
For the figure to reflect only Kodus, we recommend the API key you give Kodus is used exclusively with Kodus. If the same key also powers other tools, your provider usage — and any estimate based on it — mixes the two.

​
Configure it

The limit lives on the BYOK screen, below your Main and Fallback models.
1

Open BYOK settings

Go to app.kodus.io/organization/byok and scroll to Monthly spend limit. (The section appears once you have at least one BYOK model configured.)
2

Check the model prices

Each configured model shows its per-token price — input, output, cache read, cache write — as $ / 1M tokens. A Catalog badge means Kodus found the price automatically; a Manual badge means you’ve entered it yourself.
3

Fix any model with no price

A model badged No price couldn’t be priced automatically (a custom or uncatalogued model). Type its per-token prices from your provider’s pricing page. You can’t enable the limit until every model has a price — spend can’t be tracked for a model Kodus can’t price.
4

Set the monthly limit and enable

Enter your cap in US$, flip Enable spend alerts on, and click Save spend limit.

​
Model pricing

Kodus looks each model up in a public pricing catalog and pre-fills what it finds (Catalog). If the catalog is wrong or missing a model, type the correct per-token rates — the badge switches to Manual and your values are used instead.
On a model you’ve overridden, click Revert to catalog price. The inputs reset to the catalog values and the badge flips back to Catalog. From then on that model tracks the live catalog price again (no frozen snapshot).
Custom endpoints, self-hosted models, or models the catalog hasn’t indexed have no automatic price. Enter the rates manually to make them trackable. Until then, the limit can’t be enabled.

​
Alerts

Once enabled, Kodus checks your month-to-date spend roughly once an hour and notifies your organization’s Owners in the notification center as you cross each threshold:
ThresholdWhat you get
50%Heads-up alert
75%Heads-up alert
90%Heads-up alert
100%“Limit reached” alert
Over 100%One final “you’re over your limit — we won’t notify again this month” notice, then silence for the rest of the month
Each threshold fires at most once per month, so you won’t get repeat pings for the same level. The slate clears automatically when the new month starts.
Reviews continue running normally the entire time — at 50%, at 100%, and beyond. The alerts are the only thing that happens.
You can manage where these land (and which roles receive them) under Spend Limit in Notification settings.

​
Accuracy

Spend is an estimate. It’s derived from the tokens your reviews used and the per-token prices shown on the BYOK screen — it won’t match your provider invoice to the cent.

Keep prices accurate

Check the catalog prices against your provider’s pricing page and correct any that are off. Subscription/flat-rate plans especially won’t match per-token catalog pricing.

Use the key only with Kodus

A dedicated key means the usage Kodus sees is the usage Kodus caused — nothing else inflates the figure.
Concurrent reviews can briefly overshoot the cap. Many PRs reviewed at once near the limit can push month-to-date spend a bit past 100% before the next hourly check catches it. This is expected — another reason the real cap belongs at your provider.

​
Frequently asked questions

No. It’s notification-only. To stop spend, set a hard limit in your model provider’s billing dashboard.
Only token usage on your BYOK models, priced at the current rates shown on the BYOK screen. Kodus-hosted models don’t count.
On the 1st of each calendar month (UTC). Month-to-date spend goes back to $0 and all threshold alerts re-arm automatically.
No. Spend is tracked per model: each model’s tokens are priced at that model’s rate. If you spent $50 on model A and then switch to model B, model A’s usage stays at $50 — switching to B never re-prices A’s tokens at B’s rate. Your total just grows by what model B then uses.The only thing that re-prices already-incurred usage is a change to that same model’s price — when the pricing catalog updates a model’s rate, or you edit its manual price on the BYOK screen, that model’s month-to-date usage is recomputed at the new rate (there’s no frozen snapshot).
Every configured model must have a price (catalog or manual) and the monthly limit must be greater than $0. Fix any No price model and set a positive amount, then save.
Your organization’s Owners, in the in-app notification center. Adjust routing under Spend Limit in Notification settings.