Claude Code Basics Lesson 37 of 46
Teams: When Agents Talk to Each Other
- Sub-agents work in isolation: they don't talk to each other.
- In a team, on the other hand, they coordinate, over a shared task list.
- It's the most impressive thing to watch and the most expensive to use.
- Ninety percent of the time, a well-instructed helper is enough.
The limit of helpers
You can send several helpers in parallel, and this happens often: three rooms, three jobs, three results coming back to you. But those three have no idea the others exist.
That's completely fine for independent jobs. It becomes a problem when the second one would need to know what the first one found: in that case either you act as the bridge yourself, or the three jobs step on each other's toes.
The team: when they talk to each other
A team of agents is the level above: several agents working together by coordinating. There's a shared task list, each one picks up its own, and they pass information while working — one waits for another to finish its piece, one checks what another has done.
The difference in one line: sub-agents are three people in three closed rooms; a team is a meeting.
There's nothing to prepare in advance: you just ask out loud, saying you want a group of agents to take on that job, and it invents the roles itself depending on the task.
The bill, which is the important thing here
If a helper costs something because it starts from zero, five helpers talking to each other cost five times as much, plus coordination. And coordination isn't free: every exchange between two members is work someone pays for.
It's worth being blunt, because it's the thing that saves you the most money in this basics course: a team is spectacular to watch — seeing the figures talk to each other really does feel like science fiction — and it's almost always overkill.
When it genuinely makes sense
The job is big and splits well into pieces that can proceed in parallel.
The pieces need to talk to each other: if they're independent, ordinary helpers are enough.
You already have a clear plan in hand. A group launched at a vague goal produces five different interpretations instead of one.
If even one of these is missing, a well-instructed helper does the same job for a fraction of the cost, with far fewer things that can go wrong.
What stays in your hands
If you do try it, there's a single control point and it's at the start: the task list. It's what every member will follow, so that's where you check whether the job was understood correctly — exactly like reading a plan.
If the list is right, the rest tends to follow. If the list is wrong, you have five agents getting it wrong together, fast, and on your bill.
And now
Third bucket closed. In the next lesson we put it to the test on the most common case of all, and you'll see that the difference isn't just tidiness: it's the quality of the result.
Did you get it?
Two minutes, no grade. Trying to answer — even getting it wrong — cements the concepts far more than rereading. Not in the mood? Skip it: the lesson still counts as complete.
What is the difference between sub-agents in parallel and a team?
Two sub-agents in parallel are two strangers who know nothing of each other. In a team, on the other hand, there's coordination: they pass information and split tasks from a shared list.
When does a team actually make sense?
It's the most spectacular thing to watch and the most expensive to use. In the vast majority of cases a well-instructed helper does the same job for a fraction of the cost — and with fewer things that can go wrong.
Course Claude Code Basics -
Chapter 1 Introduction
-
Chapter 2 What's Underneath
-
Chapter 3 You Choose the Engine
-
Chapter 4 First Time at the Controls
-
Chapter 5 What You Let It Touch
-
Chapter 6 What It Knows About You
-
Chapter 7 Skills
-
Chapter 8 How Many You Put to Work
-
Chapter 9 What It's Connected To
-
Chapter 10 Automating the Setup
-
Chapter 11 Wrap-Up