From conversation to proof to running software.
The line, station by station: what humans decide, what agents build, and the proof every requirement carries when it leaves the line. Then what it means for you and your clients.
Every requirement leaves the line with its proof attached.
A factory is a sequence you can write down, repeat and check. Product Printer runs that sequence for software: humans decide what to build, coding agents build it, and every acceptance criterion is bound to a real test run before anything is called done.
- 01
Discovery
The brief, constraints and decisions on record.
Conversation, not a form
- 02
Requirements
Acceptance criteria written before a line of code.
One criterion, one check
- 03
Open questions
Nothing is built on an assumption.
Asked, answered, closed
- 04
Solution + UI
Screens and data model, reviewed.
Design decided before code
- 05
Worksteps
Sized work orders, one agent each.
Agents build, not guess
- 06
Proof
A real test run bound to each acceptance criterion.
Proof, not a checkbox
The differentiator - 07
Running software
Released, versioned, owned by you.
Yours from day one
After launch the line loops: the client starts a change from their own space (AI-assisted), the agency approves it, agents build it, the proof runs, v1.1 is released.
Requirement, criteria, workstep, pull request, test run: one row.
When a client says “that wasn’t agreed”, you open this view.
| Req | Requirement | Status | Workstep | Pull request | Test run | Criteria | Gate |
|---|---|---|---|---|---|---|---|
| FR-01 | Capture every order from every channel | accepted | WS-01 | PR #231 | #412 | 6/6 | ✓ gated |
| FR-02 | Sync confirmed orders to the ERP | accepted | WS-02 | PR #234 | #418 | 4/4 | ✓ gated |
| FR-03 | Invoice on delivery, per shipment | accepted | WS-03 | PR #237 | #421 | 3/5 | checks running |
| FR-04 | Remind unpaid invoices | accepted | WS-04 | — | — | 0/3 | queued |
| OQ-02 | Partial delivery: invoice per shipment or per order? | answered | → FR-03 | — | — | — | closed |
| NFR-01 | Hosted in the EU only | accepted | design | PR #229 | #405 | 2/2 | ✓ gated |
A recording of a real run goes here once we have one. Until then: no video, and no mock-up passed off as one.
One layer between your work and every AI model.
The Product Printer Orchestrator (PP Orchestrator) is the layer that runs every AI job of an instance — coding agents and AI agents — on the models you choose, under your rules, on infrastructure you control.
An abstraction layer
Coding agents on frontier models today, AI agents on local models when you want. Switch models — your subscription, your API keys or local models — without changing your pipeline, your projects or your client apps.
A business protection layer
For the clients you build for:
- Keys and model access stay with you: encrypted to your PP Orchestrator, never seen by the instance.
- Client code and data stay where you decide: your machine, your server, or our managed server.
- Every AI call goes through policy: routing by sensitivity, local-first options, and a hard cap on the AI credit we include in a trial.
- Every job is recorded and auditable.
Connects out, nothing connects in
Your PP Orchestrator opens every connection itself. Nothing has to reach into the machine it runs on, so it can sit behind your firewall.
Your printer. Your clients. Your revenue.
Run your PP Orchestrator on your own machine with your own subscription, or let us run it. Give every client an online space under your brand. Own the client relationship end to end, with the support to make it a great journey for them and a lasting revenue stream for you.
- Models
- yours, local or by key
- Code
- in your own GitHub organisation
- Costs
- your subscription, your rate
- Running v1.0
- Client requests a change · AI-assisted
- Agency approves · scope + price
- Agents build
- Proof
- Released v1.1
AI on your own terms.
Run your PP Orchestrator on your machine with your own subscription, or let us run it. Your models, your costs. The code goes to your own GitHub organisation, and with local models no AI call leaves your premises.
Give every client a real online experience.
Each client gets their own space: requirements, open questions, progress, change requests and documentation, all in one place. They serve themselves while it is being built: answer open questions, review designs, see the proof per criterion, request changes.
Own it end to end.
Your brand and your client relationship. Your clients see your brand, not ours.
Turn projects into recurring revenue.
Each project gets its monthly price and each change request its own price; the invoices are prepared in your own Stripe account for you to send. After launch your clients start changes themselves, with AI help; you approve scope and price; agents build; proof, then release.
Supported all the way.
Supported by the team that runs it daily.
Off the line, still running.
AvoDev, our own agency in Berlin, runs its client projects on Product Printer. Several of its client products were delivered on the factory and are still running.
An agency runs its client projects on it.
Every change gated: tests, security scans, AI security review.
The generated source and the data belong to the client.