Free Guide · No. 02

Managed Agents on the Claude Platform

A plain-language walkthrough of what Managed Agents are, how to build one entirely in the Claude Console, and what to watch out for. No code, no SDK, just the UI.

“You bring the goal and the tools. Anthropic provides the brain, the sandbox, the storage, and the dashboard.”

Send the PDF to my inbox

Drop your name and email, the guide downloads in 1 second and we'll keep you posted on new content.

PDF · 17 pages · ~30-minute build · Beginner level

What's inside

🧱

The five building blocks

Agent, session, environment, vault, memory. The whole platform fits into a five-word vocabulary.

🖼️

Mental model in one picture

Recipe, kitchen, conversation, key drawer, journal. If you forget the rest, remember this.

🪄

Quickstart wizard, end to end

Use the Console's Quickstart to ship a real research agent through the UI in 30 minutes.

🔐

Vault credentials done right

Store API keys so the agent can use them without ever seeing them. Prompt-injection proof by design.

📔

Memory that survives sessions

Attach a memory store so the agent remembers user preferences and lessons across runs.

🔧

Debug, verify, watch costs

Read the dashboard, troubleshoot first-time errors, and know what to check before you commit.

Who is this for

Curious builders

You have heard about agents and want to ship a real one without spinning up infrastructure.

Solo operators & PMs

You want a long-running, autonomous agent in production but no quarter to build the harness.

Anyone hitting API limits

Your agent runs for minutes, touches secrets, or browses the web. Time for the managed runtime.

What you'll know by the end

→What a Managed Agent is and when to reach for one
→The five-word vocabulary: agent, session, environment, vault, memory
→How to use the Console's Quickstart to ship your first agent
→How to add a vault credential so the agent never sees your keys
→How to attach a memory store so the agent remembers across runs
→How to read the dashboard, debug errors, and watch your costs
→The honest pros, cons, and trade-offs before you commit

Ready to ship your first Managed Agent?

Download the guide, open the Console, and have a real research agent running in about 30 minutes. No code, no SDK, no DevOps.

Send the PDF to my inbox

Drop your name and email, the guide downloads in 1 second and we'll keep you posted on new content.

Want the full agent playbook?

This guide gets you to a working hosted agent. AI at Work Academy takes you further: building agents in Claude Code, connecting your real work tools, scheduling, and the patterns that keep agents reliable in production. Module 1 is free.

Start Module 1 Free →