Set agent execution limits

Purpose

Kian does not want one runaway request to eat the shop's plan. Control how an agent delivers answers and put hard limits on its runs, model calls and tool calls.

Before you start

The custom agent must belong to your current workspace, and you need permission to manage agents. Larger limits and retries can use more of your plan's usage and never add capacity to it.

Steps

1. Open "Agents", find your agent under "My agents" and select it. In the window, open "Agent actions" (the three dots) and choose "Edit agent".
2. Open "Execution". Set "Stream responses" (deliver the answer piece by piece), "Timeout" (milliseconds), "Maximum steps" and "Concurrent runs".
3. Open "Model". Under "Model middleware", turn on "Model retry" and set "Maximum model retries" (0 to 10 retries after the first attempt). Turn on "Model fallback" to add up to four alternative models, tried in order when the primary model fails; the primary model cannot be its own fallback. Set "Maximum model calls per run" under "Model call limit".
4. Open "Plugins". Under "Tool middleware", turn on "Tool retry" with "Maximum tool retries" (0 to 10), set "Maximum tool calls per run" under "Tool call limit", and optionally turn on "LLM tool selector" with "Maximum selected tools" (1 to 50), which narrows the enabled tools to the most relevant ones before each model call. This limits the tools chosen by the selector. Essential tools may remain available in addition to that number.
5. Select "Save".

Expected result

A run stops with an error when it passes a hard limit. When more runs start than "Concurrent runs" allows, the extra runs wait for a free slot. Concurrent runs count all users of the agent together.

Troubleshooting