Spotify and Microsoft introduce new controls for AI agents

Spotify has introduced Xirp, an orchestration layer designed to manage AI agents within corporate environments. To prevent technical errors caused by AI agents ignoring existing infrastructure or architectural decisions, Xirp provides a centralized context for development teams. The platform is model-agnostic, allowing developers to switch between different AI engines, such as Claude or Gemini, without losing session history. Additionally, Xirp utilizes Git worktrees to allow up to 50 AI agents to operate in parallel on the same repository without interfering with one another.

In a separate development, Microsoft is adding a human-in-the-loop feature to Copilot Studio. This update will allow organizations to require explicit user approval before an AI agent executes specific actions, such as making payments, modifying files, or sending communications. This native control mechanism aims to mitigate risks associated with AI errors or malicious prompts by ensuring that sensitive tasks are paused until a human provides authorization.