The operating system we run the company on.
Takat OS is where our work lives: every client, project, task, decision and roadmap in one database, with AI assistants the team talks to in chat and on WhatsApp. We built it for ourselves — and the same building blocks go into the agents we build for clients.
System sheet
- Type
- Internal platform, built by Takat
- Doors
- Web portal, team chat, WhatsApp, each person's Claude
- Data
- One database, on infrastructure we operate
- Rule
- Assistants report only what really happened
One place for the whole company
Clients, projects, code
Client → project → repository. GitHub says what exists; the OS says who it serves and where it stands, with a health view per project.
One work list
Todos, tasks, blockers, decisions, questions, follow-ups, ideas and feedback live in one table. Closing an item means writing what was actually done.
Roadmaps
Vision, milestones, features and steps per project, with the tasks attached to each node.
Assistants in the team chat
Specialist bots create and close work, record decisions and answer “what is open on this project?” in plain words.
A WhatsApp door
The team sends notes, voice memos and files on WhatsApp; they are filed into the right project, and people are nudged about what is asked of them.
A connector for every Claude
Each teammate connects their own Claude to the OS through an MCP connector: it reads the projects, records work and proposes code on the person's branch. Every call is logged.
A feedback widget
One script tag on any site we build sends a user's feedback straight to the right project.
Starting a project in minutes
One request creates the repository, the person's branch, the project record and a first live preview.
Trust is designed in
Truth first
An assistant says “done”, “sent” or “created” only when a system confirmed it. Its reply is rewritten from the real results.
Humans send
Assistants draft; a person sends anything that leaves the company — to a client, a prospect or the press.
Nothing is deleted by a machine
A deletion request puts the record on hold and asks a human, who confirms with a code. Everything else is archived, never erased.
An AI meeting notetaker that keeps the audio in house
A notetaker joins the video meetings found in our calendars, records the audio and, after the meeting, transcribes the full recording with an open-source speech model on a machine we run — the audio never goes to an outside transcription service. The minutes land in the team chat and the shared drive.
Our first version taught us a practical lesson: a bot that waits in a meeting lobby is easy to forget. Now the organiser is told on WhatsApp or in chat the moment it waits, and a meeting the bot was never admitted to leaves nothing behind.
The same parts go into your agents
The truth filter, governed deletion, per-call logs and human handover were built and tested on ourselves first. They are what we put into the WhatsApp AI agents and platforms we build for clients, and the delivery model is described on how we build.
Frequently asked questions
Is Takat OS for sale?
Takat OS is our internal platform. Its building blocks — assistants that only report confirmed results, governed deletion, audit logs, WhatsApp and chat doors — are reused in the products we build for clients.
Where does the meeting notetaker transcribe the audio?
On a machine Takat operates, after the meeting, with an open-source speech model. The recording is not sent to an outside transcription service.
Can an assistant delete data in Takat OS?
No. A deletion request puts the record on hold and asks a human to confirm with a code; until then nothing is erased.