Why did my AI response stop early?

An AI response can stop before the task is finished because it reached an output limit or an agent step limit. Check the notice in the chat to see which limit applies.

Agent step limit

If you see Reached the 30-step limit (or another step count), the agent used all the model rounds allowed for that request. Each round can include several tool calls, such as searches or file lookups. The limit can stop a task before the model writes its final answer.

The default is 30 steps. This safeguard prevents endless tool loops, which can become expensive and degrade performance. Your chat may have a different saved limit.

To allow longer tasks:

1. Open the chat's inspector pane.

2. Find Agent Loop Control and increase Steps.

Changing the limit applies to future requests; it does not resume the stopped run. There is no Continue button for this limit. Review any completed tool actions before retrying a task, since a retry may repeat them.

Output limit

If BoltAI shows Continue generating, the provider reported that the response reached its output limit. This can happen when Max Tokens is too low or the answer exceeds the model's maximum output length.

Click Continue generating to send a visible Continue message in the same chat and request the next part. BoltAI keeps the partial answer.

For longer answers, increase or reset Max Tokens in the settings used by that chat, or ask for the answer in shorter sections. Increasing the agent step limit does not increase the model's output limit.

Other interruptions

Connection failures, provider errors, and safety filters can also interrupt a response. Follow the error shown in the chat. Raising either limit will not resolve those problems.

If the cause is unclear, contact BoltAI support with the provider and model name and a screenshot of the notice. Do not include your API key.