Two new features land for Claude Managed Agents: scheduled deployments let you run tasks on a recurring schedule, and a new environment variable credential type exposes vault secrets to CLIs without handing them to the model.
A new environment variable credential type lets Managed Agents authenticate with CLIs, SDKs, or APIs without exposing secrets to Claude. Credentials are injected at the network boundary, never visible to the model.
Each Managed Agent deployment is defined by an agent, an environment, an initial message, and a schedule. Every run starts a fresh session with no scheduler infrastructure to build or maintain.
The built-in /claude-api skill lets Claude Code set up a Managed Agent deployment on your behalf, using the ant CLI as the interface to the API.
Claude Corps will place 1,000 paid fellows full-time at nonprofits across the US, training them to use Claude effectively and helping host organizations advance their missions.
