New in Claude Managed Agents: run agents on a schedule and store environment variables in vaults

· Source: Claude Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

Claude Managed Agents now offer two new features in public beta: scheduled deployments and secure environment variable storage in vaults. Scheduled deployments allow agents to run automatically on a cron schedule, handling routine tasks like nightly data syncs, weekly compliance scans, or daily digests without requiring users to build or host a separate scheduler. Companies like Rakuten, Actively AI, and Ando are already using this for automated reporting, refreshing search answers, and managing follow-ups. Additionally, agents can now securely access CLI tools and authenticated services by storing environment variables in vaults. This enables agents to make authenticated API calls via CLIs, with the actual API key attached at the network boundary only for approved domains, enhancing security. Notion, Browserbase, KERNEL, and Milana utilize this for secure file uploads, web navigation, database interaction, and bug fixing.

Key takeaway

For AI Engineers or MLOps teams deploying autonomous agents, these new Claude Managed Agents features simplify operational overhead and enhance security. You can now automate recurring agent tasks like data analysis or compliance checks using scheduled deployments, eliminating the need for custom scheduling infrastructure. Furthermore, securely integrate external CLIs and authenticated services by utilizing environment variables in vaults, ensuring sensitive API keys are never exposed to the agent directly. This allows for robust, secure automation and broader agent capabilities.

Key insights

Claude Managed Agents now automate tasks via scheduling and securely integrate external tools using vaulted environment variables.

Principles

Method

Agents can be deployed with a cron schedule for recurring execution. For external tool authentication, register API keys in vaults with environment variable names and allowed domains; the agent's sandbox uses a placeholder, and the real key is attached at the network boundary.

In practice

Topics

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Engineer, MLOps Engineer, AI Architect

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Claude Blog.