What counts as a credit
| Event | Credits |
|---|---|
| AI reply on a channel | 1 |
| AI-drafted reply you edit and send | 1 |
| Tool call (Shopify order lookup, etc.) | 1 |
| Flow Message node sending a fixed template | 0 |
| Flow Condition / Action / Resolve / Handoff nodes | 0 |
| Knowledge ingestion (crawl, file parse) | 0 |
| Manual reply typed by a teammate | 0 |
See your usage
Sidebar → Settings → Plan & billing shows:- Current period usage vs cap
- Rolling 30-day chart
- Breakdown by agent and channel
What happens at the cap
- At 100% — new AI replies are blocked until you top up or next cycle starts. Human teammates can still reply; flows still fire condition/action/handoff/resolve nodes; knowledge still works.
- At cap + grace (Business+) — Keloa sends a warning but keeps the AI running for 7 days so you’re not stuck. You’ll be billed for the overage.
Top up
From the usage page, click Top up credits. Pick a bundle (Business+) or upgrade to a bigger plan. Top-ups apply instantly.Per-agent quota (Scale)
On the Scale plan, you can set a per-agent monthly quota to prevent one agent from consuming everything. Configure in the agent’s edit screen → Limits.Credit-saving tips
- Refine your system prompt to keep replies short.
- Use flows for deterministic paths (after-hours, tagging) — those are free.
- Use macros for the 10 most common human replies — no AI cost at all.
- Tighten knowledge — irrelevant passages make the model work harder and waste tokens.
- Disable unused tools per agent — fewer surface-area, fewer accidental calls.