Port AI enforces a hard 50-iteration cap on tool calls across all AI surfaces, including AI nodes in workflows, AI agents, and any other AI-powered interactions. For orgs on BYOLLM, where Port has no cost exposure, this cap functions as an arbitrary constraint rather than a meaningful guardrail, blocking legitimate high-context use cases like PR review agents that need to explore a larger number of files to produce safe, accurate output. When the cap is hit, execution fails.

Proposed solution: Expose the tool-call iteration limit as a configurable setting in org AI settings. Keep 50 as the default and hard cap for orgs on the Port-managed AI provider; allow orgs on BYOLLM to raise or remove the cap at their discretion.