my-mini-team

Compose your own agent workflows, run them on any task, and share them. The CLI is mmt.

Stop doing the work. Architect the team that does it.

The idea in 30 seconds

my-mini-team is three stacked layers: native skills (SKILL.md capabilities) plug into agents (roles with a model and a default skill set), and agents compose into teams — mmt's own, proprietary layer — as ordered steps and loops. Instead of prompting one agent over and over, you describe a team once and then run that team on any task. mmt run <team> spawns the team's agents as in-session Claude subagents on your subscription — via the /mmt skill, not claude -p — and you watch every step.

mmt run <team> "add SMS reminders to booking confirmations"

A team is just a small, readable file. You compose it by talking, reuse it by name, and share it so others can install it in one command.

Why it's different

• Your way, encoded. A team captures how you ship — plan, code, review, test — as a reusable object, not a one-off prompt.
• Legible. mmt show team <team> prints the whole workflow: who does what, which skills, which loops.
• On your subscription. Runs execute as sub-agents inside an interactive Claude Code session — no separate API billing.
• Shareable. mmt export <team> <dir> writes a reviewable directory bundle. Hand the dir to someone who runs mmt import <dir>.

What you'll learn here

1. Getting started — install and run your first team in two minutes.
2. Concepts — the model: team → members / loops → skills, and the team lead.
3. Create your first team — compose, refine, and run your own.
4. Skills, Loops & gates, Running, Sharing.
5. Reference — every command and the team-file format.

New here? Start with Getting started →