Somebody has to build the systems a company runs on. We already did.

Before we sold any of this, we needed it ourselves — a chief of staff that never forgets a meeting, a first ninety days that lands as a working plan instead of a deck, a way of building interfaces our own agents cannot drift from. We built it, we run our own ventures on it, and now you can have it three ways: take the pieces into what you are already running, hand us the job and direct it, or have the whole thing installed in your company with your own people running it afterward.

No pricing on this page. Engagements are scoped and priced once we understand the work.

However you want to run it, this is how you'd get it.

This is not a pricing ladder, and nothing is held back from the smaller option. It's three different answers to one question: a year from now, who is the one actually running this day to day — you, us, or your own team once we've handed it over?

Parts

You already have engineers, and you would rather own this than depend on us every time it changes. Take the pieces — the design language, the gate that enforces it — into your own codebase, and they are yours from there.

Your team runs it. We hand it off and step back.

Operated

You do not want to hire for this yet, or at all. Tell us what it is for, and the actual work happens on our side — you review what comes back and correct course, the way you would direct anyone doing this job for you.

We run it day to day. You steer it.

Factory install

You want this to become something your own people run, not something you keep paying someone else for. We install the same machinery inside your environment and train your team until the room no longer needs us in it.

Your team runs the whole factory once we leave, not just what it made.

Pick a mode. Watch the week change.

Not three tiers of the same product — three different Mondays. Choose P, A or F below and the week underneath it changes to show who is actually doing the work each day.

IllustrativeOne composite week per mode, not a client's real calendar — the same three modes above, walked day by day instead of stated in one paragraph.

Parts — You run it

  1. MonYou

    You pull the pieces into your own repo — the design language and the gate that enforces it.

  2. TueYou

    Your engineers wire it into your build. Nobody on our side is in the loop.

  3. WedYou

    A change ships that breaks the pattern. Your own CI catches it, the same way ours does.

  4. ThuYou

    You extend it — a new component, a new token — because it's yours now.

  5. FriYou

    Nothing to report back to us. You already own the whole thing.

What's included in Parts
  • The design language and its enforcement gate, handed off into your own codebase
  • Everything needed to run and extend it yourselves, with nothing left behind that only we can touch

We do not operate anything on your behalf in this mode — no ongoing support beyond the handoff itself.

Operated — We run it, you steer

  1. MonYou

    You tell us what needs attention this week. We start the actual work.

  2. TueDana

    We run it day to day — the briefing, the document, the build — on our side.

  3. WedYou

    Something comes back that needs your call. You review it and correct course.

  4. ThuDana

    We adjust based on what you said and keep going, without a second meeting.

  5. FriDana

    You steered the week. We did the work underneath it.

What's included in Operated
  • The day-to-day work itself — briefings, documents, builds, checks — run on our side
  • A review point built into the week, not a status call you have to schedule

The system stays on our side while we operate it. If you want to own the instance outright, that is mode F.

Factory install — Your team runs it after we leave

  1. MonDana

    We install the machinery inside your environment, next to your own people.

  2. TueYour team

    Your team watches how it runs, then starts running pieces of it themselves.

  3. WedYour team

    We hand off the parts your team has learned; our support tapers as they take over.

  4. ThuYour team

    Your team runs the day on their own. We're in the room only if asked.

  5. FriYour team

    The room no longer needs us in it. Your team owns the whole factory.

What's included in Factory install
  • The same machinery installed inside your own environment
  • Training that continues until your own team runs it without us in the room

This is a handoff with an end date, not an ongoing service — once it is installed and trained, we step out.

Here's the actual machinery, doing actual work.

Not hypothetical services — five things we run our own company on right now, including where the proof is still thin. Read them as evidence for what the top of this page just claimed, not as a menu to pick from.

A chief of staff that remembers

It knows your calendar, your people, and what happened in your last three meetings, and it surfaces the one exception that matters instead of a feed you have to read. You do not onboard it with a survey — you talk to it, and it stops asking questions it should already know the answer to.

Operated — we run it
We provision it and run it behind the scenes. You use it: read the briefing, answer what it raises, tell it things.
Install — your team runs it
We install the whole stack in your own environment and hand your team the keys. You own the instance and everything in it.

Private beta. Provisioning is a person doing the work rather than a signup page, and there is no self-serve path in either mode.

A first ninety days that is a working system

An incoming executive — a new CEO, a founder taking over a function, a leader walking into a company mid-story — gets one working artifact instead of a deck: the meetings and the decisions and owners inside them, the relationship terrain around the seat, and a narrative that holds it together. It lands before the ninety days start.

Operated — we run it
We assemble it. Your part is access to the calendar and the meetings, and confirming who and what actually matters. What changes hands is the working artifact itself.

Assembled from four things we run on ourselves daily, not built new for the occasion. No engagement in this shape is published, so read this as a described capability rather than a demonstrated one.

A demo, before anything is priced

Tell us what is stuck and we read what is public — your product, your site, your docs. What comes back is a reply from a person, not a pitch. This is the entry point for every operated engagement, which is why it comes before a number rather than after one.

Operated — we run it
We read it and reply. Your part is to exist as a company with something public to read, and to answer anything we ask.

A person reads every one of these, and replies within three business days.

A look your product can't drift away from

Your product keeps drifting off its own look, because every builder — including the agents doing some of the work — makes a slightly different small call. This closes the gap: one shared visual language, plus a machine check that stops a change before it ships if it breaks the pattern. Not a style guide anyone can ignore — a build that fails.

Parts — you run it
Take the shared visual language and its safety check into your own codebase and wire them up yourselves.
Operated — we run it
We adapt it to your brand, build it into your product, connect the safety check to your release process, and ship it live.
Install — your team runs it
We install the safety check in your release pipeline and train your team to run and extend it, then step out.

One shared visual language, enforced automatically, on the system's own product. This site is not built on that system yet — the design system page says so too.

A track that actually survives being played

A good take is not a finished track — it still has to hold up on a phone speaker, sound right collapsed to one channel, never clip, and land at the tempo it is supposed to. This is the check that confirms all of it, run against a file that is locked down first, so the numbers can never drift from the audio they describe.

Parts — you run it
Bring a finished render, run the fixed check, read the results. Same file, same result, whoever runs it — there is no listening session to schedule.
Operated — we run it
We carry the whole production to that standard and lock the delivered file against later drift. You direct the sound; none of the technical work lands on your side.

Nothing here links to audio. The check is numeric and a stranger can verify it; the material it has run on is not ours to publish.

Your AI wrote it. Here's how you know it's true.

Every AI-written claim gets labeled the moment it is written — sourced, estimated, proposed, or unproven — so a fact and a guess never look the same twice. The page carries a live version of the same check: paste your own text and watch it get labeled in your browser, nothing sent anywhere.

Parts — you run it
Take the labeling taxonomy, the verification gate, and the drift-detection hooks into your own repo and vault.
Operated — we run it
We run the verification loop and labeling protocol against your own outputs and hand back a labeled record.
Install — your team runs it
We wire all four components into your existing repo and workflow, and train your team to run and extend them.

Proposal awaiting sign-off before it ships to production — read as a described and demonstrated mechanism, not yet a sold engagement.

A ledger of who decided what while your agents did the work

When the board asks "who decided that," this is the receipt — every human touchpoint in an agent-run engagement, classified and appended as it happens, with every delivery registered against it. The page carries a real, redacted 58-touchpoint timeline you can scrub through yourself.

Parts — you run it
Take the taxonomy starter set and the append discipline into your own vault or repo.
Operated — we run it
We keep the ledger on your engagement and hand you the same two artifacts we produced for ourselves.
Install — your team runs it
We map your own decision-boundary classes and install the append discipline into your tooling.

Proposal awaiting sign-off before it ships to production. No named buyer has adopted this yet — what exists is the record of it running on our own build.

One door, whichever mode you want

Parts, operated or installed — the first step is the same, and it is not a call. You send a company URL and one sentence about what is stuck, and what comes back is a reply from a person. Engagements are priced per engagement and scoped once we understand the work.

A person reads every one of these, and replies within three business days.

We reply here and nowhere else.

Whatever is most public. A landing page is fine.

One sentence. The specific thing, not the category. Say which mode you have in mind if you already know.

A person reads every one of these.

No calendar link, no discovery call, no sequence. The next thing you get from us is a reply.

We store what you send here to reply to it. We do not add you to a list, because we do not have one. The full detail is on the privacy page.

See how a start works →

Three fair questions

What does it cost?
Priced per engagement and scoped once we understand the work. There are no bands on this site, because the number depends on what the work turns out to be and we do not know that before we have read you. You see it before you have committed to anything.
Which of these can I actually have?
All five, in the modes listed on each card. None of this is self-serve — every mode starts with a person reading what you sent.
Can I take one piece and leave the rest?
Yes, where a card lists Parts. They are separable because we needed them separable ourselves — the design language is enforced in CI against its own repository, and the release gate returns the same numbers on the same file whoever runs it.

Whichever mode you want, it starts with a demo.

We read every request — three business days, either way.