A Charter — the constitution
Budget caps per corporation, per day, and per mission. Prohibited actions. Risk and reversibility ceilings. Escalation rules naming real humans. The kill-switch policy. Signed by you; enforced by the platform.
zero AI expertise required
early access — founder-delivered
Don't hire a company. Charter one.
This page is for the person who wants the outcome of a company without becoming an AI engineer to get it. You bring the business and the judgment; The Agentic Corporation brings the company — a full, autonomous AI-agent corporation that does the day-to-day work, under strict built-in guardrails. It is the closest thing to a money printer you can responsibly build: you pick the business and the strategy (those are real business decisions), and we take the guesswork out of running it. Free to start.
Honest hedge: "money printer" is a metaphor, not a promise of income — you choose how you make money; we provide the frontier execution. Questions? Email a human — a person replies: aipeteaipete@gmail.com
step 1 — the zero-knowledge how
Pick an archetype, not an org chart
You do not configure agents, prompts, or models. You pick a work-shape — a ready-made starting corporation — and we wire the frontier scaffolding underneath it. Archetypes are work-shapes, not industries: pick the one that matches how your work flows, then deepen it with marketplace skills for your domain. The platform already carries worked examples in finance, healthcare, legal, and manufacturing.
Why this matters: safe, reliable, production-grade agent autonomy is rare — most agent projects never reach it. Gartner expects over 40% of agentic-AI projects to be canceled by 2027, with only a small fraction running in production today (see how we build for it). The hard part is making autonomous agents do real work without going off the rails; taking that off your plate is the whole point of starting here.
| Archetype | Work-shape | Roles you get |
|---|---|---|
| research | Retrieve → extract claims → adversarially verify → synthesize cited reports | orchestrator · analyst · fact-checker · security officer |
| ops | Triage telemetry → score deterministically → draft runbooks and containment plans (humans execute) | orchestrator · telemetry analyst · runbook engineer · security officer |
| build | Draft typed design specs → review and risk-rank changes → produce verification evidence (humans merge) | orchestrator · design engineer · reviewer · security officer |
Every archetype ships with a security officer role and separation of powers built in: no agent reviews its own output, and the riskiest steps stay human.
step 2
Name it. Scaffold it. Sign it.
$ agentcorp init my-corp --archetype ops scaffolded corps/my-corp charter: budgets · prohibited actions · escalation · kill-switch policy roles: 4 (deny-by-default tool grants) missions: 2 examples, budget-capped next: replace placeholders, then a human signs the charter
Scaffolding takes minutes. Going live is deliberate by design: a human signs the
Charter (an unsigned one fails closed and will not run), and every tool grant is an explicit,
capped decision. There is no --allow-all flag — the platform refuses to mint one.
That deliberateness is not friction; it is why you can sleep while your corporation does the
work that earns its keep.
what you get
A whole company in the box
Roles with separation of powers
An orchestrator to plan, specialists to do, a reviewer to check, and a security officer to object — different agents, so nobody grades their own homework. Roles that read untrusted content hold no privileges; privileged roles never read untrusted content directly.
Missions with budgets
Work arrives as missions: scoped goals with their own budget envelope nested inside the Charter's caps. The platform debits the budget before every model and tool call and refuses at zero — a runaway agent stops at the number you signed, not at your card limit.
An audit trail and a kill-switch
Every event lands in a hash-chained, tamper-evident ledger — who did what, under which grant, at what cost. And if you ever want everything to stop, the kill-switch is yours: named humans can trip it; agents cannot untrip it.
in developmentA live "spaceport" dashboard — your corporation rendered as a working scene with roster, missions, spend, and alarms — is being built as the customer app alongside the platform. Here is the whole company in the box, made tangible:
the cost story
Near-zero to start. Usage-priced to run.
A corporation has no salaries, no offices, no minimum spend. Every plan runs exactly one corporation and you pay a platform base fee plus metered usage — model tokens and tool calls, debited against ceilings you control. There is no per-seat billing; human admins are a membership allowance, not a billed seat. When your corporation is idle it is designed to scale to zero, so an idle company costs approximately nothing in compute. And because the budget pre-flight is fail-closed, your worst month is the cap you set, not a surprise.
- Start free on Launch ($0/mo): a real, full corporation — one corporation, an elastic frontier agentic workforce, and a starter usage cap, with the complete security baseline. It is not a crippled trial; nothing core is held back to be unlocked later.
- Climb the plans when you want a larger frontier agentic workforce (more worker replicas), a higher concurrent-mission ceiling, and more frontier intelligence — never a corp count, never a billed seat, all the way up to large corporate scale. See the full plans on the pricing page.
- Buy capability, not consultants: add marketplace skills — signed, verified, privilege-bounded — instead of hiring for every specialty.
who's in charge
You are. Structurally.
Founding an agent corporation does not mean trusting agents with your name. The Charter you sign is the ceiling no agent can raise: agents cannot expand their own grants, budgets, or policies — that boundary is enforced in the architecture, with changes to it requiring two humans. Escalations default to "stop and ask you", timeouts mean deny, and the riskiest actions can be configured to always require your approval.
who's behind this
A named founder, delivering by hand
The Agentic Corporation is built by Peter Vu Nguyen, its founder. There is no anonymous "we" here and no team to hide behind: this is early, founder-delivered work, built in the open. The mission is simple and the reason the whole thing exists: giving money-making AI-agent businesses back to displaced workers — a zero-knowledge, security-first home where one person, with no AI expertise, can charter a reliable corporation that does the work that makes money. You make the business calls; we remove the guesswork of running it.
What you see described here is the architecture as it exists in the open repository — no customers, metrics, or certifications are claimed. Right now you work directly with the founder, and a working corporation is set up with you; this is early access, not a stranger self-serving a finished product this minute. That is the point: you get a person who replies, not a ticket queue. If you want to start a corporation or just ask a question, email a human — a person replies: aipeteaipete@gmail.com.