Claude Code, Agent Teams, and paid ads ops

How I Run Paid Ads Like a Team of 5 (As a One-Person Company) Using Agent Teams in Claude Code

I’m Hunter, founder and CEO of Obieo. Most “marketing work” isn’t marketing. It’s the messy operations around marketing: landing pages, tracking QA, reporting, follow-up, and all the glue that gets rebuilt every time you launch.

Agent Teams (an experimental Claude Code feature) is the first thing I’ve used that actually feels like having a small performance team in the room. Not for vibes. For output.

Want this set up done-for-you?

I’ll help you set up Claude Code + automations for your paid ads workflow. Start here: Automate with Obieo

What Agent Teams actually does

Agent Teams lets you coordinate multiple Claude Code instances on the same codebase. One session plays lead. Others work as specialists. They message each other, share a task list with dependencies, and avoid stepping on the same files.

The difference from subagents is communication. Subagents report back to a main session. Agent Teams can coordinate peer-to-peer without you acting as the router.

Subagents are contractors you send on errands. Agent Teams feels like a project team sitting in the same room.

System requirements and setup

1. Enable Agent Teams

Add an environment variable. Agent Teams is experimental and off by default.

// ~/.claude/settings.json
{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  }
}

Or export it in your shell:

export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

2. Optional: split panes

Agent Teams works in a single terminal, but split panes make it easier to watch teammates work.

brew install tmux

3. Verify access

claude

Do you have access to agent teams?

The use cases that actually win for paid ads

Agent Teams is more expensive in tokens, so I only use it when parallelization buys back real hours.

1. Creative sprint + landing page build

One teammate explores angles and hooks while another builds the landing page and a third audits tracking. The output is a coherent “ad promise → page proof → measurement” chain.

2. Tracking and attribution triage

When performance tanks, you need competing hypotheses in parallel: tracking break vs landing page issue vs audience drift vs creative fatigue.

3. Reporting and alerts that don’t waste your week

A teammate can wire automated reporting and drift alerts while others ship the next iteration. This is where “one person company” starts feeling real.

The prompt template I actually use

If you want a starting point, this is the prompt I use when I’m building marketing systems around paid ads.

Create an Agent Team for paid ads execution for [offer] targeting [audience] on [platform].

Spawn: angle-strategist, copy-variants, landing-page-builder, tracking-qa, ops-automation.

Output:
1) angles + hooks
2) ad copy variants grouped by angle
3) landing page updates in repo
4) tracking checklist + fixes
5) lightweight reporting/alert automation plan

Coordination rule: teammates message each other directly to resolve mismatches (promise, proof, tracking).

If you want this done-for-you

If you’re a founder, marketer, or agency operator and you want Claude Code + automations wired into your workflow, I’ll help you set it up and make it usable on day one.

Start here: https://www.obieo.com/automate