<aside> 📌 Claude just dropped "Managed Agents" — and it changes everything. Here's exactly how to set one up, connect your tools, and let it work while you sleep.

</aside>

What Are Managed Agents?

Before this update, if you wanted an AI agent doing work for you — answering emails, updating your CRM, monitoring your ads — your laptop had to stay open. Close the lid? Agent dies. Lose wifi? Done.

Managed Agents fix that. Your agent now runs on Anthropic's servers. Not your computer. Not some janky VPS you set up at 2am. Their cloud. Their infrastructure. Your agent.

• It runs while you sleep
• It runs while you travel
• It runs with your laptop completely off
• It picks up right where it left off — files, memory, everything persists

How It Works (3 Pieces)

You don't need to be technical to understand this. There are three things:

• Agent Config — This is the brain. You tell it what model to use, what it's allowed to do, and what tools it can access. You set this up once and reuse it forever.
• Environment — This is the workspace. Think of it like giving your agent a computer in the cloud with everything pre-installed. Python, Node, whatever it needs — already there.
• Session — This is the agent actually running. You start a session, give it a task, and it goes. The session stays alive, remembers your files, remembers your conversation. It doesn't reset every time.

That's it. Config → Environment → Session → Agent is live.

Connect Any Tool Without API Keys

This is the part that's genuinely exciting. Before, connecting your agent to tools like Gmail, GitHub, Slack, or your CRM meant copying API keys, pasting them into config files, and praying nothing broke.

Now it uses OAuth. Same thing you do when you click "Sign in with Google." One click, you're connected. No secret keys floating around in your code.

• Connect Gmail — your agent reads and responds to emails
• Connect Slack — it posts updates and answers questions
• Connect GitHub — it monitors repos and opens pull requests