Anthropic shipped an update for Claude Managed Agents yesterday that brings three features many have been waiting for. Managed Agents — agents created and operated persistently via the API — now get scheduling, secure secrets, and browser access.
Cron-Based Scheduling
Managed Agents can now run automatically on a cron schedule. You define the schedule, and the agent executes at the specified time — no manual trigger needed.
Sounds simple, but it’s a big step. Until now, developers had to build their own orchestration to run agents on a regular basis. Now it’s native to the platform.
Vaults for Environment Variables
The second feature is Vaults — secure storage for environment variables. API keys, credentials, and other secrets can now be stored encrypted without living in the agent’s code.
Notion is already using Vaults to securely run CLI tools alongside MCP servers. That shows enterprise customers need this kind of security infrastructure before they’ll deploy agents in production.
Browser Capabilities
Through an integration with Browserbase and KERNEL, Managed Agents can now navigate the web. They can load pages, fill out forms, and extract information — all programmatically without human intervention.
This significantly expands what agents can do. An agent that regularly compares prices, fills out forms, or monitors websites previously required custom infrastructure.
My take: This update shows where things are headed. Managed Agents are evolving from an interesting experiment into a production-ready tool. Scheduling plus secure secrets plus browser access — those are exactly the three building blocks you need to let an agent work truly autonomously. Anthropic is methodically building the infrastructure that enterprise customers demand.
Sources: Anthropic Blog, Releasebot