FeatureFebruary 25, 20261 min read
Agent Loops
Agents can now run multi-turn reasoning loops across API, chat, and MCP with configurable turn limits, reflection, cost budgets, execution history, and telemetry.
agentsloopsautonomymcp
Agent loops bring multi-turn execution to API, chat, and MCP surfaces. You can set maxTurns, enable reflection, cap cost, and persist iteration history instead of stopping after a single response.
This makes it practical to ship research, review, and support agents that can plan, act, inspect results, and continue until they reach a real stopping condition.