# October 15, 2026 - Human-in-the-loop approval policy for tool calls **URL:** \ **Category:** Application Enablement\ **Tags:** app-enablement, Feature, ai-agents\ **Created:** [October 21, 2026, 7:14am UTC](https://community.cumulocity.com/t/october-15-2026-human-in-the-loop-approval-policy-for-tool-calls/15276 "2026-10-21T07:14:32Z")\ **Posts on this page:** 1\ **Page:** 1
**Author:** ![cumulocity](https://dub1.discourse-cdn.com/flex005/user_avatar/community.cumulocity.com/cumulocity/32/31990_2.png) [@cumulocity](https://community.cumulocity.com/u/cumulocity)\ **Post date:** [October 21, 2026, 7:14am UTC](https://community.cumulocity.com/t/october-15-2026-human-in-the-loop-approval-policy-for-tool-calls/15276/1 "2026-10-21T07:14:32Z")
## Context * * * **Change Type:** Feature **Product area:** Application enablement & solutions **Component:** AI Agents **Deployed at:** [eu.latest.cumulocity.com](http://eu.latest.cumulocity.com), [apj.cumulocity.com](http://apj.cumulocity.com), [jp.cumulocity.com](http://jp.cumulocity.com), [cumulocity.com](http://cumulocity.com), [us.cumulocity.com](http://us.cumulocity.com) > **Technical details** > > **Build artifact:** ai-agents (2.3.0) > **Internal ID:** MTM-67219 ## Description * * * When using AI-powered features that make tool calls (such as API requests or system commands), you need to ensure that sensitive operations receive proper oversight. Previously, there was no mechanism to enforce approval policies for these tool calls, which could allow unintended or risky operations to execute without human review. Now, the system evaluates and enforces human-in-the-loop approval policies for tool calls, ensuring that operations requiring oversight are reviewed and approved by authorized users before execution. This change impacts all users and applications that use AI features with tool calling capabilities, as tool calls will now be subject to configured approval policies, providing better control and security over automated operations in your installation.