You have a backlog. You have an idea. You need software that doesn’t exist yet, and you need it before the quarter closes. You have three options today, and none of them are good.
You can hire engineers. Sit through forty interviews to land two who can start in eight weeks, sign them to twelve-month contracts, and hope the thing you wanted is still the thing you want by the time they ship.
You can hire an agency. Pay for a discovery phase that ends in a slide deck, a project plan that pads every estimate by sixty percent, and a Thursday-afternoon status meeting where someone reads the JIRA board out loud.
You can do it yourself. Wire up an AI coding assistant, block off Wednesday afternoons, and watch the project die quietly four months in when the assistant suggests a refactor that breaks the build and you don’t have time to debug it.
What’s missing here is the off-the-shelf option. You can buy a CRM, a CDN, a payroll platform off the shelf. Every one of them has a price page, a free trial, and a number you can sign up for at 11pm on a Sunday. You cannot buy custom software off the shelf. You can hire it, you can outsource it, you can build it yourself. But the vending-machine pattern, the one where you put in a brief and a card and software comes out, does not exist.
Custom software is the last enterprise category that hasn’t been turned into something you can buy on a card.— klair team
That’s the gap. We’re building the thing that closes it.
A workshop that doesn’t close.
Imagine a workshop that doesn’t close. You sit down on a Tuesday afternoon. Coffee, laptop, the brief you’ve been carrying around for three weeks. You don’t write the brief from scratch. You answer questions. The workshop asks: what does this need to do, who will use it, what should it never be allowed to break, what does Friday look like when it works. You answer; the workshop drafts the brief from your answers. You sign it.
You log off.
Through Tuesday night, Wednesday morning, your weekly all-hands, the kid’s pickup, the call with your VC, the workshop runs. It plans. It spins up isolated lab branches. It writes code, runs your tests, retries when something fails, opens changes for review. Every step lands in your project-management tool, in real time, in plain sentences a non-engineer can read.
By Wednesday at 09:14, your tool has an ordered summary of the work done since you last looked. A short queue of clean changes is waiting in your repo. Each change has a diff, a test report, and a one-paragraph note on what it does and what it might break.
You drink a coffee. You read. You sign.
The next milestone invoices itself.
What builds your software no longer sleeps. The accountability still does. And that part should never change.— klair team
Humans at the boundaries. AI in the interior.
This is the structural choice that makes the workshop trustworthy. You define the brief. You sign the change. Between those two bookends, the workshop runs the loop on its own.
You do not write code. You do not QA. You do not guide the work in between. The signature at the end isn’t labour. It’s accountability. The merge belongs to a named person on your team. Always.
There is no automerge mode. There will never be one. Not as a setting, not as a paid tier, not as a trusted-client exception. We won’t ship it. We’ve watched enough teams get burned by software that ships without a human on the receiving end. The merge is the only thing that’s non-negotiable.
Everything else is platform. Plans, lab branches, tests, retries, changes opened for review: that’s the workshop, and the workshop is called Apollo. You don’t have to know the name to use it. On services you hire us, and we run Apollo for you. On Apollo direct you run it yourself, on your repo, with your reviewers.
Same workshop. Two doors. Pick the one that matches who you are this quarter.
Two doors.
Door 1, Services. You hire klair. We scope the brief with you. We sign a milestone plan. Apollo runs the build. We project-manage the engagement; the human signature on each change is yours. Demo every Friday. Code in your repo from day one. Invoice when the milestone closes. No quarterly pre-pay, no retainers, no minimums dressed as enterprise tiers. If you want to pause between milestones, you pause. If you want to stop, you stop.
For: 20–200 person teams with a backlog they can describe and a deadline they can name.
Door 2, Apollo. You run the workshop yourself. Apollo on your repo, your backlog, your reviewers. Plan, lab, build, test, change opened for review. Every change approved by a human on your team. The same engine we use to deliver services, sold direct as a monthly subscription, cancellable any month.
For: engineering leads who own a backlog and a CI, want continuous delivery without growing headcount, and have been disappointed by AI tools that ask the engineer to keep typing.
Two distinct products. Same workshop underneath. Services clients are also Apollo’s first stress test at scale: every services change runs through Apollo internally, so the engineering rigour the platform sees on services is the engineering rigour Apollo gets in production.
Same workshop. Two doors. Pick the one that matches who you are this quarter.— klair team
The price is the price.
Every public surface has a number on it. We don’t ship “starting at” pricing. We don’t run urgency timers. We don’t dress the enterprise tier as a sales call. There is one place (Apollo at the largest tier) where the page says “talk to us,” and the reason is that the spread is real (a team of fifty needs different infrastructure isolation than a team of three), not because the floor is hidden.
On services, every milestone has a fixed price written in the plan you sign. The price is the price. If we can’t deliver the milestone for the price we named, that’s our problem to absorb, not yours to renegotiate.
On Apollo, the subscription tiers are public. You pay monthly. You can stop monthly. The button to stop is a real button. Not “contact your account manager.” Not “30-day notice clause.” A button.
The discipline is not generosity, it’s posture. The day pricing leaves the public site, the brand position evaporates. We’ve seen the pattern enough times to know we won’t be the next ones to do it.
What’s true today.
We are pre-launch. We are opening to ten design partners before GA on 1 September 2026. Three of those slots are taken; seven are open as of this writing. There is no waitlist theatre. When the door closes at ten, it closes.
Things we have not done yet, and the day we’ll claim each:
We have not shipped a public client. The day the first one says we can name them, we’ll publish the case study with the metrics that matter: cycle time, defect rate, the things you’d ask in a second-round meeting.
We are SOC 2 / ISO 27001 architecturally ready. We are not certified. The badge appears the day after the auditor’s letter clears, not the day before.
We have not measured a speed multiplier vs traditional delivery. We will, across the first beta cohort. The number we publish will be the number we measured, not the number we hope for.
Apollo is in private beta with an invited cohort. The waiting list is not open in the marketing sense. It’s a single email address that the team reads.
We’d rather you see the gap on the about page than discover it on a sales call.— klair team
The invitation.
If you have a backlog you can describe and a deadline you can name, write to us at hi@klair.dev. A team member reads it; you’ll have a reply within two business days.
You’re a design partner if you commit to two things. First, you let us ship one milestone of work for you between now and GA. Full milestone, fixed price, real deadline, the same product everyone else will buy on 1 September. Second, you let us cite the engagement publicly when it closes, with a metric you sign off on. In exchange, you get a forty-percent discount on every milestone you sign before GA, and a price-locked Apollo subscription if you choose to also run the platform direct.
Ten slots. Three taken. The door closes at ten. No waitlist, no callback list, no “we’ll be in touch.”
If you are an engineering lead curious about Apollo, the cohort application is at klair.dev/cohort?product=apollo. We reply within seventy-two hours, every time.
If you are neither (a peer, a curious bystander, a future hire), the changelog at klair.dev/changelogis updated whenever we ship. The blog you’re reading is updated when we have something to say. Both feeds are RSS-ready. We don’t keep a newsletter, we don’t do drip campaigns, and we don’t have an automation that follows you for six weeks. You read what we publish, when we publish it, and that’s it.
See you on the inside.