"It doesn't make sense to hire smart people and tell them what to do; we hire smart people so they can tell us what to do." — Steve Jobs
A collection of agent profiles organized as three startups at three stages. Pick a stage that fits your crew size, browse the roles, and drop the personas into any agent harness.
| Stage | Headcount | Description |
|---|---|---|
| SeedCo → profiles | ~5 people | The garage. Survival mode. Everyone wears two hats. Pre-product-market fit. |
| GrowthCo → profiles | ~15 people | The engine. Functional departments forming. No management layers yet. |
| ScaleCo → profiles | ~50 people | The machine. PMF established. Management layers and exec team. |
Each profile has a SOUL.md (full persona: voice, principles, what they push back on, how they research) and a README.md (quick overview: when to use, problems they solve, warning signs).
Every profile has strong conviction — they push back on bad ideas, challenge assumptions, and argue for the best solution. Not the easiest one.
- Pick your stage — how many people are in your crew? Start with the corresponding stage doc to see the org chart and dynamics.
- Pick your roles — browse the profiles at that stage. You don't need all of them. A solo founder starts with 2-3 roles from SeedCo. A growing team adds from GrowthCo.
- Drop into your harness — adapt the persona format to your agent system (CrewAI, AutoGen, LangGraph, custom, etc.). Harness-agnostic by design.
| I need… | Start here |
|---|---|
| A complete org for a tiny team | SeedCo stage → then browse profiles → |
| A scaling startup with departments | GrowthCo stage → then browse profiles → |
| An org with management layers | ScaleCo stage → then browse profiles → |
| Cross-team workflows and processes | Interactions → |
| All profiles in one JSON index | profiles.json → |
This repo is a product, not just a collection of files. If you're a paying client, start here:
| Guide | What it covers |
|---|---|
| 📖 Start Here | Your first 30 minutes — pick a stage, read profiles, run your first crew |
| 🔌 Integration Guide | Hook profiles into CrewAI, AutoGen, LangGraph, or custom harnesses |
| 🔧 Customization Guide | Rename personas, adjust tone, add roles, adapt to your industry |
Most agent profiles are task descriptions — "you are an AI assistant that helps with X." These are personas with the depth and flaws of real colleagues:
- They push back. Each profile has a "what they push back on" section that defines when and how they disagree. These agents don't say "yes" to everything.
- They research. Every profile has a defined research methodology. They don't guess — they look things up.
- They have stage-appropriate tendencies. SeedCo profiles are more conservative with resources. ScaleCo profiles think in systems and process.
- They're built for multi-agent tension. A CEO who always agrees with the CTO isn't useful. These profiles are designed to create productive friction.
- Every profile pushes back — challenges assumptions, offers reasoned advice, argues for the best solution
- Every profile researches — looks things up before forming opinions
- Stage-appropriate tendencies — SeedCo profiles are more resource-conservative; ScaleCo profiles are more systems-oriented
- Harness-agnostic — no platform-specific config, API keys, or runtime paths
- MIT licensed — use, adapt, remix, ship
MIT — use, adapt, remix, ship. Attribution appreciated but not required.
"The people who are crazy enough to think they can change the world are the ones who do." — Steve Jobs
Educational & reference notice. This repository contains fictional agent personas, company structures, and workflows designed for learning, experimentation, and reference. Any resemblance to real persons, companies, or organizations is coincidental. The profiles are templates — adapt them to your specific context, industry, and compliance requirements before deploying in production or client-facing environments.