Bring your own API key for the live chat bot
AI bot usage is billed based on tokens consumed through your own API key. Deskwoot does not charge per message.
This article covers bringing your own API key (BYOK) for the live chat bot, Fynn. With BYOK you connect your own OpenAI or Anthropic key, those bot conversations are billed by your provider at their rates, and they do not count toward your Deskwoot AI conversations meter. BYOK is available on the Startup plan and above, and it applies to the live chat bot only.
If you are looking for how the standard meter works (the included AI conversations, overage rates, and auto-recharge), that is a separate topic. See AI conversations meter, overage, and auto-recharge.
What does bring your own key cover?
BYOK applies to Fynn on your website live chat only. It does not change how Copilot actions (suggest reply, summarize, translate, detect language) are billed, and it does not apply to bots on other channels. When BYOK is on for a live chat bot:
- Fynn calls your provider with your key, so you pay your provider directly for tokens.
- Those live chat bot conversations are excluded from your Deskwoot AI conversations meter.
- You pick the model your key has access to, so you control cost and capability.
How do I connect my key?
- Go to Settings > Bots and open the live chat bot you want to run on your own key.
- Choose your provider, OpenAI or Anthropic.
- Paste your API key and save.
- Confirm the key is valid and has quota with your provider, then send a test message to the widget.
How is BYOK usage billed?
Your provider bills you by tokens at their published rates. A token is the basic unit AI models use to process text, and every message Fynn reads and every reply it writes consumes some. Token usage goes up with longer questions and answers, more conversation history sent for context, and more capable (pricier) models. Deskwoot adds no markup on your provider's rates, and these conversations stay off the AI conversations meter.
Keeping costs in check
- Keep Fynn concise. Shorter answers use fewer tokens and often read better.
- Set spending alerts with your provider. Both OpenAI and Anthropic let you cap usage or get billing alerts.
- Trim the context. Avoid sending unnecessarily long histories with each request.
- Watch your provider dashboard. Your provider's billing screen is the source of truth for exact charges.
Troubleshooting
- Fynn stops replying. Check that your key is still valid and that your provider account has funds or quota left.
- Costs higher than expected. Long conversations, a traffic spike, or verbose instructions can all raise token use. Review your provider dashboard.
- Invalid key. Re-paste the full key with no extra spaces and confirm it is active on your provider account.