My agent made a tool call with a bunch of arguments but didn't put ending curly braces, leading to nothing actually being passed into our function. Call ID: 019c6e10-91e4-7112-a00a-cf7a32127a7f Pay attention to the very last tool call in this call. There are no ending curly braces, causing it to error out. Is there anything I can do to always ensure the arguments are properly formatted when being sent into the tool call?

can u send what arguments u r sending and what u get?

Yes, here are the defined arguments, and the arguments the chatbot passed in, without the ending curly braces https://cdn.discordapp.com/attachments/1473532239361998889/1473722541750222940/image.png?ex=69973e92&is=6995ed12&hm=2a33db8534bf130721ed4ef1ee8d985755a6164ac2592da5e0da428c74136b3e& https://cdn.discordapp.com/attachments/1473532239361998889/1473722542098223174/image.png?ex=69973e92&is=6995ed12&hm=beb9748ed3a9ea8645e89f322b9f2dccf20042534a05319121966d86527b110c&

Hi, Thanks for sharing the call ID. We reviewed 019c6e10-91e4-7112-a00a-cf7a32127a7f and identified the issue. The final tool call was cut off mid JSON because your assistant’s maxTokens is set to 250. The model generated speech plus a 9 argument tool call in the same turn, exceeded the 250 token limit, and the output was truncated before the closing curly braces. Since the JSON was incomplete, it parsed as {} and your server correctly rejected it. Fix: Increase your model’s maxTokens to at least 1000. You can update it via API:

{
  "model": {
    "provider": "anthropic",
    "model": "claude-sonnet-4-5-20250929",
    "maxTokens": 1000
  }
}

Or go to Dashboard → Assistant → Model → Max Tokens and increase it there. 1000 gives enough headroom for larger tool calls plus speech. You can also remove the cap and use the model default. After updating, test the appointment flow again and confirm the full arguments are received.

Awesome, thanks! Will increasing the max tokens increase the time it takes for the assistant to make responses?

Increasing the max tokens may lead to slightly longer response times, as it allows the assistant to generate more content in one go. However, the difference should be minimal.