Skip to content
Let’s talk

Case study · Cloud machines for AI coding agents · USA and worldwide

detachboxA founder, a fractional CTO and one engineer took detachbox from an idea to a commercial product in three weeks

detachbox gives Claude Code, Codex and other CLI agents an always-on Linux box: the developer closes the laptop and the agent keeps working, on the Claude or Codex subscription the developer already pays for. We built it as founder → fractional CTO → engineer, with no team in between, and it went from the founder's idea to paid plans in three weeks. I own the architecture, the infrastructure, the code review and the technical direction, and we keep developing it. No subscription token or API key ever lands in a box, which makes detachbox one of the safest ways to run remote agents on your own subscription.

Get the full case as a PDFAll case studies
Client
detachbox
Product
Always-on boxes for AI coding agents
Model
Your own Claude or Codex subscription
Main market
USA, plus the rest of the world
My role
Fractional CTO and startup advisor
Team
Founder, 1 engineer and me
detachbox: the product

In numbers

from the founder's idea to a commercial product with paid plans
3 weeks
subscription refresh tokens or API keys inside a box, even with sudo
0
agents preinstalled: Claude Code, Codex, OpenCode and Gemini CLI
4
to roll a box back to a snapshot
2–6 s
a month for a box that never sleeps, where Codespaces left on costs about $135
$19
for the limits and reset times of every Claude and Codex subscription
1 screen

The product

The public site: the always-on box with Claude Code at work, the limits board across subscriptions, every agent tab with the ones that need an answer on top, and where everything runs. The product screens use sample data.

detachbox.com
  • The Accounts board: every Claude and Codex subscription with its 5-hour and weekly limit, reset time and the boxes using it
  • Terminals: every agent tab across all boxes, the ones that need an answer on top, next to a phone with notification settings
  • Architecture: the browser, phone or SSH reach the box through the dashboard, SSH or private previews, and the token vault gives the box an access token only
01

The starting point

Coding agents now run for hours, but on a laptop they stop the moment the lid closes, and a bare server means a day of setup and keys left on its disk. detachbox gives every developer an always-on Linux box, ready at login, with the agents preinstalled and the developer's own Claude or Codex subscription connected. The founder needed a commercial product, not a prototype, and needed it fast.

We built it as founder → fractional CTO → engineer. The founder owns the product and the market; I own the architecture, the infrastructure, the review and the technical direction; one engineer builds. There is no product team, no QA department and no ops staff between us.

02

What we did

  1. Three weeks from the idea to a product on sale

    The founder brought the idea, I turned it into an architecture and a plan one engineer could build, and I reviewed the work as it landed. Three weeks later detachbox was a commercial product with three paid plans, Stripe billing and a refund promise on the first overnight run. A team hired the usual way spends that long just staffing up.

  2. The subscription the developer already pays for

    Each box comes with Claude Code, Codex, OpenCode, Gemini CLI, Docker and a full developer toolchain preinstalled. The developer connects the Claude or Codex subscription they already have, so they pay detachbox for the machine and not a second time for the model, and model traffic goes straight to Anthropic or OpenAI without passing through detachbox.

  3. Keys never enter the box

    I designed the security model around one rule: an agent running with sudo must hold nothing worth stealing. Subscription refresh tokens stay on detachbox's servers, encrypted with AES-256-GCM, and a box receives only a short-lived access token renewed on schedule. API keys for other services stay in a vault built on Sallyport Cloud, a credential gateway for agents I also architect, so the agent makes the call and never sees the key. Disconnecting an account deletes its tokens from the server and from every box that used it.

  4. Boxes walled off from each other and from the platform

    Each box is an unprivileged system container with its own user ID range, capped CPU, memory, disk and process count, and Docker running inside it. Private networks, neighbouring boxes, detachbox's own servers and outbound mail are blocked. Previews run on a separate domain, so a dev server inside a box can never read the dashboard's cookies. SSH takes keys only, and sign-in uses passkeys.

  5. Infrastructure that grows node by node

    I run the infrastructure. Boxes live on detachbox's own hardware, and capacity grows by adding nodes, not by redesigning anything. A box's files and Docker volumes survive a node reboot, the kernel is patched weekly, and a snapshot rolls a box back in 2 to 6 seconds. Owning the hardware is what keeps an always-on box at $19 a month, where GitHub Codespaces left on around the clock costs about $135.

  6. Built for agents that run all day, not for watching them

    Every agent tab across every box shows whether it is working, needs the developer or is done, and a push, Telegram or email notification arrives when an agent needs an answer, which can be given from a phone. One limits board shows every Claude and Codex subscription's usage and reset time, and when one runs out, the box moves to the subscription with the most room left. Every port of a box gets its own private HTTPS address for previews.

03

The result

A founder, one engineer and a fractional CTO took detachbox from an idea to a commercial product in three weeks, and it keeps growing on the same scheme. A developer closes the laptop and the agents keep working, on the developer's own subscription, in a box where no long-lived token or API key ever lands. That security model, on infrastructure that scales node by node, makes detachbox one of the safest ways to run remote AI agents today.

04

What comes next

We keep developing detachbox the same way: the founder sets the product priorities, I keep the architecture, the infrastructure and the review, and the engineer ships. Business plans already add snapshots, priority support and company invoices, and Enterprise adds dedicated nodes and an SLA.

oleg.isSent on request

Full case · PDF

detachbox

A founder, a fractional CTO and one engineer took detachbox from an idea to a commercial product in three weeks

  1. 01The architecture and the three-week plan, week by week
  2. 02The security model: tokens, the vault and box isolation
  3. 03How bring-your-own-subscription works and where the tokens live
  4. 04The infrastructure: nodes, snapshots and how capacity grows
  5. 05How the founder, the fractional CTO and the engineer split the work
  6. 06What we would do the same way again
Prepared by Oleg SotnikovPDF

The full detachbox case

The public story stops here. The PDF has the rest: the architecture before and after, the migration plan, how the team works with AI agents, what it all costs to run and where the savings came from.

I send every case myself and use your email for nothing else.

More case studies

All case studies
01AulaHousing and utilities · KazakhstanFractional CTOThree releases a day instead of one in four months, 3 people doing the work of 16, and a cloud bill down 70%Aula is one app for apartment residents and the 109 companies that manage their buildings, with 247,000 accounts. As fractional CTO I rebuilt how the product is made and run: development and testing went from 16 people to 3, the servers from 18 to 3, the cloud bill fell by 70% and errors in the mobile app by 97%. A task now reaches production in a day instead of 2 to 3 weeks, and the product ships three times a day at 99.99% uptime, a pace most in-house teams never reach.Read the case study03CV RocketJob search and AI CVs · USA and worldwideFractional CTOOne engineer and a fractional CTO took CV Rocket from zero to sales in five weeks and to 1,000+ customersCV Rocket writes a CV for one job posting in 15 to 50 minutes and brings the US and world job market into one place: more than 1.5 million postings from 89,000+ job sites. I designed the architecture, the cloud services and the payment integration, and one engineer and I built the product in four weeks, a launch that usually takes a full team several quarters. It sold from week five and now has more than 1,000 customers.Read the case study04MetizPromServiceMetalworking and parts manufacturing · KazakhstanFractional CTOA factory saves $300,000+ a year and halved its IT costs after two engineers I hired replaced its contractorsMetizPromService machines parts to customers' drawings in Kazakhstan. When I joined, contractors ran its IT and its accounting, production and lathe-control systems, and the company paid for outside software and licences. I built its own infrastructure, hired and trained two engineers, and we replaced every paid production system with an ERP and CRM of our own, with the ERP live in 3 months. The company now saves more than $300,000 a year on software, spends over 50% less on IT than the contractors cost, salaries included, and its CNC machines stand idle 23% less, with half as many unplanned repairs.Read the case study

Want the next case study to be about your company?

In a 30-minute call we pick the first task to hand to AI and estimate what it will save you.

Book a callThe call is free, and you talk to me directly.