@Jorden Williams I'm still not really clear on how this works. Do I need to create an Open API account and add money to that wallet? I noticed that when I did not do that, the bot failed (this was all on my own website). I saw the rebilling video, but the part that's still confusing. Does Assistable track the chat time and rebills the client; which shows up as a credit in my Stripe account and from there I would pay my Open API account? Or do I need to ask the customer to create their own Open API account and top it off when balance goes low?

On August 26, in just one hour, our API Key consumed $29.84, which projects to $21,480 per client per month if that rate were to continue. To put this in perspective: - With 10 clients, this would scale to $214,800 per month. - With Assistable, the cost would only be $3,000 per month. - And the correct cost using GPT-5 mini directly should be just $200 per month. This clearly demonstrates a serious error in token accounting: instead of the expected ~8.5k tokens per conversation, the system is logging around 22k tokens, inflating consumption by more than 1,000x. It is urgent to verify whether the system prompt is being duplicated or if the wrapper/API is sending redundant data. ⚠️ This is a projected analysis assuming each client consumes 15,000 conversations with a prompt of ~8,500 tokens per conversation. We urgently need help with this, because if the API Key were billed correctly, the cost would align with expected consumption. Even compared with Assistable’s API Key pricing, the current difference is abnormally high. I hope I have explained my point clearly. @Jorden Williams

Is there like a new client onboarding SOP? I'm looking for forms or docs that will help me get the information I need to do the prompting and help me set up a new account.

For outbound calls, can the AI agent start the call with, “Hi, calling for ________.” for each contact? (Custom value?) Do I just add this in the prompt? Thanks!

I'm getting an error on my chat widget. Sorry, I encountered an error. Please try again. I though this was fixed.