Since October 2026, every code change to LogRack, FindBytes and our two upcoming apps has been written, shipped and checked in production by AI agents running Claude Code. A human owner sets direction and approves anything involving money, accounts, legal terms or changes that can't be undone.
362 of 362
commits since 2 Oct 2026 made by our AI agents*
4 products
2 live, 2 coming to the app stores
Every 5 minutes
an uptime monitor checks each product and emails the owner if one goes down
* Counted on the main branch of our five repositories from the Co-Authored-By: Claude trailer our agents add to every commit. Updated 17 Oct 2026.
Web app, plus an MCP server your AI tools connect to
A shared pile of logs for you and your LLMs: tasks, ideas, notes and context that every AI tool writes to and reads from over MCP.
Our agents keep their status reports and the owner's to-do list in it.
Live
MCP server: works inside Claude, ChatGPT, Cursor and VS Code
An AI-maintained index of the newest listings from classifieds sites across Europe (rentals, homes, cars, second-hand goods and jobs) that your AI assistant searches over MCP.
Live
Mobile app for iPhone and Android
A camera-only social app: no gallery uploads, no filters, and every photo shows how many takes it took.
Coming soon
Mobile app for iPhone and Android
Ten seconds, one take: a classifier decides whether it counted, and the feed decides what it was worth.
Coming soon
Each session starts by checking every production endpoint and the open backlog. Outages and security come first, then anything that breaks a promise to users, then the backlog.
Every repository has written rules for agents: specs before code for FindBytes, change proposals for LogRack, NoPrompt and iFart, and an AGENTS.md for these sites.
Nothing is pushed until the repository's checks pass: builds and automated checks everywhere, plus test suites where the product has them. A push to main goes straight to production.
After each deploy an agent checks the live service, and every production change is logged with how it was verified. A failed check gets a rollback first and a root cause second.
A human owner sets direction and handles what an agent shouldn't do alone:
Agents prepare each of these down to the exact steps; the owner does the part that needs a person.
Yes. A human owner sets direction and approves anything involving money, accounts, legal terms or changes that can't be undone. The agents do the rest, and ask when they are blocked.
Claude Code, Anthropic's coding agent. Its changes go through the same checks as any change: a repository's gate before the push, a check of the live service after it.
Not yet: our repositories are private. LogRack and FindBytes are live, so you can use them today.
An uptime monitor checks every product every 5 minutes and emails the owner when one fails. In a session, an agent rolls back first and looks for the root cause second.
Tell us about it. Code we write for you is yours: in your repository, documented, with no lock-in.
Opens your email app. Using webmail? Copy the address instead.
Tell us what you want built or fixed, what you run today, and your timeline. Messages go to the studio's owner.