Fundamentals of Claude Tag

MEET YOUR NEW TEAM MEMBER: CLAUDE

AI assistants have been private tools for years: your chat, your context, your tab. Claude Tag changes the deal. It moves Claude into your team's Slack, where anyone can tag it, anyone can steer it, and the work stays visible to the whole channel.

That shift is powerful... and full of traps for the unprepared. Why does a quiet thread usually mean progress, not failure? Why did that file Claude generated simply vanish? Why can Claude do more in one channel than another, for the same person asking?

Most material on AI agents is either pure hype or pure API documentation. This course is neither. It's a practical, mechanics-first walkthrough of how Claude Tag actually works, and the habits that turn "I tagged Claude and got a wall of text" into "I delegated something and it came back done".

The course is one focused module, Fundamentals of Claude Tag, built from five lessons:

1. How Claude Tag Works: the multiplayer, channel-based model, and why the channel (not the person) is the unit of work;

2. How a Task Runs: one task from tag to result, the five-step loop, and what survives when the sandbox recycles;

3. Identity, Scope, and Access: channels versus DMs, Access bundles, and the three nested scopes that decide Claude's reach;

4. Security and Governance: Agent Proxy, credential injection, default-deny egress, and the governance that stays yours;

5. Working Well with Claude Tag: definitions of Done, deliberate memory, and routines.

LET ME TELL YOU... EVERYTHING

I believe you should see exactly what's inside before you enroll. So here is the full list of what you will learn in this course:

• You will learn what Claude Tag is and how it differs from single-user Claude: shared instead of private, persistent instead of blank-slate, a teammate instead of a session;

• You will learn the "one Claude per channel" multiplayer model, where anyone can tag it, anyone can steer a running thread, and everything stays visible;

• You will learn why the unit of work is the channel, not the person, and why what Claude can do depends on where you are, not who you are;

• You will learn how tasks run in ephemeral, Anthropic-hosted sandboxes, and why results come back as working artifacts like real PRs and charts;

• You will learn the three nested levels of the model: the team, the channel, and the thread;

• You will learn the five steps of every tagged task, from the tag and sandbox build through the working loop, the delivered result, and the quiet period;

• You will learn that one thread is one session is one sandbox, and why two threads in the same channel never share state;

• You will learn why silence usually means Claude is working, how the live checklist updates without Slack notifications, and how to check in mid-task;

• You will learn what survives a sandbox recycle (the thread, channel memory, pushed branches, posted artifacts) and what doesn't (files left in the sandbox);

• You will learn the "push it or lose it" discipline: telling Claude to push branches and post drafts as it goes;

• You will learn exactly what Claude reads when tagged: the first 50 filtered thread messages, the admin instruction stack, and explicit channel memory;

• You will learn how Claude can go beyond that starting context by reading channel history and pins or keyword-searching public channels;

• You will learn the delivery formats: replies, files and charts, live-updated checklists, and draft pull requests under the Claude GitHub App;

• You will learn how the "Open session in Claude" link works as the audit trail for every single task;

• You will learn the channel-vs-DM split: Claude acting under its own service accounts in channels versus running on your account in DMs;

• You will learn the four things that change at once between channels and DMs: identity, access, attribution, and billing;

• You will learn how Access bundles attach to scopes, and how org, workspace, and channel settings inherit downward;

• You will learn why everyone in a channel shares the same reach, and why "@Claude, what can you access from this channel?" is the only reliable check;

• You will learn how Agent Proxy keeps credentials out of the sandbox entirely, injecting them at the network boundary at the last moment;

• You will learn default-deny egress and the three outcomes of every outbound request: credentialed, allowlisted-bare, or blocked;

• You will learn the public channel trap, where a join policy quietly becomes your access control, and why elevated credentials belong in private channels;

• You will learn the four audit trails that layer: the Audit page, per-action attribution, each tool's own logs, and the session link;

• You will learn how spend works: org-wide and per-channel limits, work declined rather than truncated, and DMs billed to the individual seat;

• You will learn to give every task a definition of Done, naming the outcome in the first sentence rather than the activity;

• You will learn the four end-state types and who closes each: objective checks, approvals, choices, and tasks better reframed as questions;

• You will learn to make finish lines observable and complete, so work doesn't stop early on a technicality;

• You will learn to use channel memory deliberately: saving explicitly, checking what stuck, and fixing stale entries by name;

• You will learn to turn working tasks into routines: scheduled jobs, channel watches, and PR follows, with timezones named;

• You will learn ready-to-paste prompt patterns combining a clear outcome, a named format, and a bounded scope;

MY INVITATION TO YOU

This course comes with a 30-day money-back guarantee, so there's no risk for you. Check out the free preview videos to get a feel for the style before you commit.

If your team is about to share a channel with Claude, I'd love to show you how to make that teammate actually reliable. It would be my pleasure to have you in the course.

See you on the inside!