Founder OS Kit

Stop hoping your AI agent won't merge to main. Give it a constitution.

The exact governance system — approval gates, budget envelopes, kill/scale criteria, cross-session memory — that runs a real, revenue-generating one-person software company with a team of autonomous coding agents.

This is the same rulebook running Guestly and the rest of the SYNNEX portfolio, in production, today. Not a demo repo — the actual file.

The problem

Your agent got good enough to scare you a little.

The first few weeks of agentic coding feel like magic. Then you give it more autonomy — background runs, its own todo list, work across sessions — and a different feeling shows up: you stop knowing what it's about to do.

Failure mode 01

Your agent got within one command of merging straight to main — and the only thing that stopped it was luck, not a rule.

Failure mode 02

Your agent re-litigated a decision you made last week, because nothing persisted the why — only the code changed.

Failure mode 03

You have no reliable way to know, at a glance, when an agent is one step away from something you can't undo — a charge, an email, a deploy.

Failure mode 04

Ten background agents, ten half-finished threads, and no single place that says "these three things need your decision."

The solution

Governance-as-code, not another prompt library.

Five pieces, wired together, checked into git. Every agent reads them before it does anything.

01 · Constitution

CLAUDE.md

The single file every agent reads first. Company principles, tech defaults, and the hard rule: agents never merge to main.

# CLAUDE.md — company constitution ## Principles 1. The founder decides, agents execute + advise. 2. If it's not in a file and not in git — it didn't happen. 3. Cheap before pretty.
02 · Agent charters

One file per agent, per role

Scope, tools, and autonomy boundaries defined in a checked-in charter — not tribal knowledge in your head.

--- name: revenue-agent tools: Read, Grep, WebFetch scope: billing-readonly ---
03 · Approval gates

A fixed list of "always ask"

Spending money, emailing an external party, deploying to prod, deleting data, committing to a price — every agent stops and writes it down instead of doing it.

Draft blog post — SEOauto-approved
Send pricing email to lead⏳ pending you
Deploy landing to prod⏳ pending you
04 · Budget + kill/scale

Envelopes, ledgers, exit criteria

An agent with money to spend gets a pre-approved envelope and a running ledger — never a blank check. Every project ships with a written "when we stop" before it starts.

# kill-criteria.md kill_if: 0 paying sales by week 4 scale_if: 1+ sale → set MRR target
05 · The dashboard

A live control room, not a folder of markdown

React + FastAPI. Review what's pending, approve or reject from your phone, see the full reasoning behind every request — without digging through git log.

Pending approvals3
Active agents4 running
Budget envelopes this monthwithin limit

Built and running in production

Not a theory. The same system, live.

From the pending-approval log

An agent drafted a discount email to every registered user — good reasoning, real numbers, solid copy. It never sent it. The constitution routed it to a pending-approval file instead, next to the agent's own list of what it didn't know yet. A cheaper experiment got approved instead.

## APR-0XX | draft email — 50% off, all users
status: ⏳ waiting on founder — not sent, not scheduled

Pricing

One tier. One price. No subscription.

$79

Founder OS Kit — one-time payment, yours forever. No subscription: you own the files, not a login.

  • CLAUDE.md constitution template
  • 2–3 agent charter templates, ready to adapt
  • Memory + approval protocol (the exact pattern above)
  • Budget-envelope + kill/scale criteria templates
  • Onboarding + hire-agent skills
  • SECURITY.md checklist
  • Full dashboard source (React + FastAPI)
  • Deploy guide (Cloudflare Tunnel — free tier)
Get the kit — $79

Questions before you buy?

Talk to a human

Not sure it fits your setup? Ask — a real person answers.

FAQ

Questions before you buy

Does this only work with Claude Code?

The pattern is tool-agnostic — it's a set of files and rules, not a plugin. It's written and tested against Claude Code, but the same constitution, approval-gate, and memory files work with any agentic coding tool that can read a markdown file at session start (Cursor, others).

Do I need to know Python or React to run the dashboard?

No — the deploy guide walks through a copy-paste setup on Cloudflare's free tier. Knowing React/FastAPI helps if you want to customize it, but it runs as-is out of the box.

Where is my data stored?

Everything lives in your own local files and your own git repo. There's no SaaS layer and no SYNNEX server that sees your data — the dashboard reads your local files directly.

Is this a subscription?

No. One-time payment, yours to keep and modify. If we ever ship a paid add-on (like a 1:1 setup call), it'll be optional and separate.

What if it's not for me?

This is built for solo founders and small teams already running autonomous coding agents who've hit the "wait, what did it just do" stage. If you're brand new to agentic coding, save your money for now — you don't have this problem yet.

About SYNNEX

We're a one-person software house, run with an AI agent team.

SYNNEX Software Solutions builds SaaS products — events (Guestly), fintech, sports-tech, and more — with a team of one founder and a fleet of specialized AI agents. The Founder OS Kit is the exact operating layer that makes that possible: the same constitution, approval gates, and budget discipline running every product in the portfolio.

Cheap before pretty. Every solution is evaluated on monthly cost first. Free tier beats subscription, serverless beats a server.

Multi-tenant from day one. Every product is built so it can be duplicated for the next customer without touching the code.

Simple beats clever. No abstraction, library, or service gets added without a proven need.

Guestly — events Founder OS Kit — this page More in the portfolio