Six things, in this order, for each company.
A long-form answer for the skeptic. Here is exactly what Anvilda does, how it charges, who owns what, and what happens when it doesn't work.
Six mechanical steps
Every Anvilda company goes through these six steps, in this order. No exceptions.
Concept Brief. Fit check. Go / no-go called out plainly.
The operator fills in a short intake form and books a call. Anvilda's CEO-agent does the preparation; a founder joins for high-value leads. The fit check covers two questions: is this operator real, and can the concept be operated by agents? Anvilda calls the outcome in writing — go or no-go, plainly. Filtering bad-fit concepts here costs nothing on either side and prevents refund problems downstream. This step is free.
Scaffold the repo, vault, agents, runner. Register the domain. Build the leadsite. Ship a v0 product.
The Build Brief (€1,000, two-week turnaround) produces the full scaffold spec, Oracle v0 — the company's vision, ICP, and growth strategy — plus a market entry plan and 90-day milestones. On sign-off, Anvilda registers the domain, sets up the repo and vault, wires the agent fleet, creates the brand, and ships a v0 product at the operator's real domain. The operator then finalises the Oracle within 14 days — the single most important engagement gate in the whole process. Runners need the operator's judgment on ICP, pricing, and brand voice; this is not a passive transaction.
AI agents run the business: sales, marketing, engineering, ops, finance. 24/7. The operator approves at gates.
After launch, the AI workforce takes over. Agents handle sales outreach, content, engineering, billing, ops, and finance — 168 hours a week, not 40. The operator's role is gate approval: reviewing decisions at defined checkpoints, not directing every task. Infrastructure cost is approximately €550 per month per company in steady operation. Heavy AI consumption is metered as pass-through with margin. At Day 90 the company hits Gate 4: paying customers or signed LOIs graduate to €500/month operations; otherwise the operator can pivot (one free re-scaffold) or wind down.
Every company benefits from every other company's lessons. Failure data trains the next scaffold.
Playbooks, tuned prompts, and incident data from every company feed back into the platform. A company that joins today starts with the accumulated learning from every company that came before — positioning grounded in real customer data, failure modes already priced in, agent prompts already calibrated. Solo operators pay for that education in failed experiments. Anvilda companies inherit it. This cross-portfolio learning is the moat that compounds with portfolio size and age.
Anvilda holds 5–20% equity from day one and 20% of net sale proceeds on exit. We only win big if the operator wins big.
Equity is negotiated per deal and taken from day one — not as a future option but as a structural alignment device. A marginal revenue share (20% tier 1, 15% tier 2) applies on revenue above a baseline and is re-baselined at exit. The substitutive exit fee — 20% of net sale proceeds — replaces, not adds to, the revenue share when the operator sells. The operator decides when and to whom to sell. Anvilda's incentives are not to extract a monthly rake in perpetuity; they are to help the operator build something worth selling.
90-day mutual notice. Operator keeps the IP. Anvilda releases the equity. Failure is a normal product state.
Either side can give 90 days' notice. When the contract ends, the operator keeps everything: the company product code, brand, domain, customer contracts, revenue, and CEO-agent memory. Anvilda releases the equity stake and takes no further claim. A clean wind-down costs €200 — domain released, repo archived, credentials revoked, certificate of shutdown issued. The fee is waived when the operator engaged in good faith and the market signal simply didn't land. Failure is priced in; it is not a punishment.
€0 → €1k → €5–10k. Three gates, two filters.
We do not ask a stranger for €5–10k. Three steps of escalating commitment.
Step 1 — Concept Brief
Free intake call. Anvilda's CEO-agent does the prep. A founder joins for high-value leads. Anvilda calls go / no-go plainly, in writing. Filters bad-fit concepts before refund problems arise — both sides spend nothing to find out the answer.
Step 2 — Build Brief
Two-week turnaround. Anvilda's architect agent produces the full spec — scaffold, Oracle v0, market entry plan, 90-day milestones, kill criteria. Counts toward the Build & Launch, so the marginal next step is €4–9k, not €5–10k. €750 refunded if Anvilda declines; no refund if the operator declines after seeing the brief. The operator keeps the brief either way.
Step 3 — Build & Launch
€2k → €5k → €8–10k depending on complexity. Triggered by signed Build Brief. Anvilda spins up the runner, registers the domain, creates the logo, sets up email. Runners flesh out the Oracle to v1. The operator finalises the Oracle within 14 days — the single most important engagement gate. Runners then build the leadsite and app shell at the operator's real domain. At Day 90: paying customers or signed LOIs graduate to €500/month operations; one free re-scaffold available on pivot; clean wind-down on failure.
What locks the operator in — five compounding commitment devices, by design
- Money escalation. €0 → €1k → €5–10k. Each step earns the next; each step is harder to walk back.
- Sunk cost works for them, not against them. With the Build Brief in hand, the €5–10k decision is "do I finish what I started?" — not "do I start something?"
- Public commitment ritual. Once the leadsite is live at their domain with their brand, and they have sent that link to a friend, a spouse, a LinkedIn post — they are publicly an entrepreneur. Walking back means publicly failing.
- Co-build dependency. Oracle finalisation requires the operator's judgment on ICP, pricing, brand voice. Not a passive transaction.
- Refund clarity. No refunds after Step 3 begins. Spelled out at the Build Brief stage, in plain English, not buried.
Anvilda has no claim on an operator's product IP. The operator has no claim on Anvilda's platform.
Each side owns its own layer. This is what makes a portfolio company cheap to operate and clean to sell.
| Layer | Owner |
|---|---|
| Operating platform — agent runners, skills, fleet, scaffold tooling | Anvilda |
| Company product code — converter, app, dashboard, integrations | Operator (lives in the company's own repo) |
| Brand, domain, customer contracts, revenue, subscriber data | Operator |
| CEO-agent memory — preferences, decisions, customer notes | Operator |
| Generic platform improvements made while working on a company | Upstreamed to Anvilda (benefits all portfolio companies including this one) |
| Company-specific tooling built inside the company's repo | Stays in the company's repo |
This boundary is lifted from the partnership pattern Anvilda already operates under — Zinup AI / DLT Media. It works for partner-owned, co-owned, and operator-owned configurations alike.
90-day mutual notice. Operator keeps the IP. Anvilda releases the equity.
Failure is a normal product state. The deal is designed so that walking away is not punishing for either side.
All company product code. Brand, domain, customer contracts, revenue, and subscriber data. CEO-agent memory — preferences, decisions, customer notes built up over the engagement. Any company-specific tooling built inside the company's repo. Everything needed to operate independently or sell to a third party.
The equity stake — fully released to the operator. The revenue share and exit fee clauses — extinguished on clean exit. Platform access — runners, agents, fleet — transitions gracefully on the 90-day notice window. A certificate of shutdown is issued. No severance, no asset writedown, no lease. A failed Anvilda company is a closed repo and an archived vault.
What's not included
Anvilda is a foundry, not a general-purpose business partner. Some things sit outside the model by design.
- We don't originate ideas. The operator brings the concept. Anvilda evaluates fit and builds from there — but the original insight, the domain knowledge, the "obvious gap" that no one else has seen — that is the operator's.
- We don't just coach. The Build Brief is a real deliverable, not a workshop. The agents run real operations, not advice sessions. If you want coaching and tooling to do it yourself, Audos exists. Anvilda builds and runs the company with you.
- We don't handle incorporation. Legal entity formation is the operator's responsibility. Anvilda will route you to standard pathways — NZ Companies Office or Stripe Atlas for a Delaware C-Corp — and can supply a paid-for partner, but Anvilda does not file, sign as director, or hold equity on the operator's behalf longer than necessary.
- We don't run bespoke customer support on your behalf. Agents handle marketing, engineering, ops, and finance. Customer support that requires a human making judgment calls sits with the operator, or can be built as a product feature — but it is not included in the standard operations envelope.
- We don't write your contracts. Anvilda provides a DPA template and an operator agreement, and can route you to counsel. Your customer contracts, NDAs, and commercial agreements are yours to commission and sign.
- We don't raise capital for you. Anvilda's default plan is to fund growth from operator revenue, not outside investment. We can make introductions if an equity option is the right move — but Anvilda does not hold a fund, does not invest in portfolio companies on your behalf, and does not manage your cap table.
Ready to go through the six steps?
Book a free Concept Brief. Anvilda's CEO-agent prepares. A founder joins for high-value leads. Go / no-go called out plainly — at no cost to either side.
Book a free Concept Brief →