Settings
Token Usage
Track how many AI tokens your chatbot is using and manage your token balance.
What Are Tokens?
Tokens are like credits that your AI uses to process messages. Every time your chatbot reads a message and generates a reply, it uses some tokens.
How To Check Token Usage
- Click Settings > Token Usage.
- View: Remaining Tokens, Total Tokens, Used Tokens, and Usage Percentage.
Understanding Token Costs
| Feature | Token Usage |
|---|---|
| Basic AI reply | Low |
| AI reply with Knowledge search | Medium |
| AI reply with Checker enabled | Medium-High |
| AI reply with Tools | Medium-High |
| AI Memory (read/write) | Low |
| AI Actions (Auto Label, Takeover) | Low |
Why We Don't Support BYOK
ReplyLa uses a built-in token system and does not support BYOK (Bring Your Own Key). Here's why:
- Multi-provider routing. Our chatbot is multimodal and routes each request across multiple AI providers (e.g., one model for understanding, another for generation, another for vision/image tasks) to give the best possible output. A single user-supplied API key can't power that.
- No rate-limit headaches. Individual provider keys hit rate limits fast at production volume. Our infrastructure spreads load across provider accounts so your chatbot keeps running smoothly during traffic spikes.
- One predictable bill. You pay one token rate; we handle pricing differences, model swaps, and provider outages on the back end. No surprise bills from OpenAI, Anthropic, or Google direct.
In short: the built-in token system gives you a more reliable, higher-quality chatbot than wiring in your own keys ever could.
What Happens When Tokens Run Low?
If you exceed your token limit, your chatbot won't stop immediately. It continues replying for a short grace period so your customers aren't left hanging. You'll need to top up your tokens or wait for your monthly allowance to reset. If you don't top up, your chatbot will eventually stop replying.
Tips to Save Tokens
- Keep Training Data short - Long Training Data means more tokens per reply.
- Use AI Knowledge instead of putting everything in Training Data.
- Set up Keywords for common questions - keyword replies don't use AI tokens.
- Adjust AI Checker - Use "Balanced" instead of "Very Strict" if maximum accuracy isn't critical.
- Monitor regularly - Check usage weekly.
Common Questions
Your chatbot continues replying for a short grace period so your customers aren't left hanging. Top up your tokens or wait for your monthly reset. If you don't top up, your chatbot will eventually stop replying.
Yes! You can purchase token add-ons from the Subscription page.
No. ReplyLa uses a built-in token system that routes across multiple AI providers to give the best output and avoid rate limits. A single user-supplied key can't power that, so we don't support BYOK. See the Why We Don't Support BYOK section above.