Watcher is an organization-wide security layer for AI coding agents such as Claude Code and Codex, developed by Apollo Research. It deploys consistent policy across all developer devices via MDM, blocks dangerous agent actions in real time, and provides centralized session analysis to detect incidents. Teams gain full observability into every agent session, with automated analysis that flags security failures and recommends corrective rules. Watcher supports both cloud-hosted and self-hosted deployment and is built on frontier AI safety research trusted by OpenAI, Anthropic, Google DeepMind, Meta, and Microsoft.