marketing & content teams
pain: every post, newsletter and summary starts from a blank page and a busy person.
factory: brief in → drafts in your voice → human review → saved where you publish from.
> fabrik --help
fabrik-raw-flow [PRIVATE PILOT]
ai factories for your business.
fabrik raw flow turns repeatable knowledge work into ai flows that run on your own machine, with your own claude account. the ai does the work. your people decide what ships.
The problem: AI that does not repeat
ai assistants help one person at a time. the work around them stays manual: someone pastes the brief, someone checks the draft, someone forgets to publish. nothing repeats by itself, nothing is measured, and every result depends on who was at the keyboard that day.
a factory fixes that: the same input, the same stations, the same review, every time. people stop operating the ai and start approving its output.
> fabrik show --flow
How Fabrik Raw Flow works
- 0 · trigger what starts it: you, a schedule, a github issue, a file dropped in a folder, another flow.
- 1 · raw what comes in, and from whom: a brief, a recording, a spreadsheet, a link.
- 2 · flow the process, station by station. each station is one claude code call with its own rules and limits.
- 3 · review a human approves, asks for changes, or turns a comment into a permanent rule.
- 4 · destination where the result goes: a folder, a pull request, a published report.
- general manager. a chat above the map: ask how things are going, launch a flow with instructions, or describe a new flow and get its draft.
- previas. every output waits in a review queue with timing per station, usage and your decision box. one comment can become a rule.
- the tray. drop audio, images, video, csv, pdf or a link: fabrik transcribes or converts it so a flow can use it.
- fabrik server. assign fabriks to people, see usage per person and flow, get alarms by telegram or email when something breaks, publish approved reports as html pages.
> fabrik who --for
Who Fabrik Raw Flow is for
small dev teams
pain: small issues wait weeks because nobody has a free afternoon.
factory: github issue with a label → branch, tests, pull request → you merge. only the commands you allow.
operations & foundations
pain: meetings, reports and calendars live in people's heads and inboxes.
factory: recording or spreadsheet in → structured report → approved → published as html for the board.
Proof
- built and used daily by the founder: the content flow that writes his own posts, the dev workshop that opens pull requests on dunning lite.
- pilot team: a foundation with ~30 people, each on their own claude account, managed from one fabrik server.
- [PILOT METRIC: hours replaced per week · pieces approved vs sent back] shown here when measured. nothing invented.
> fabrik faq --objections
Objections, answered
"is our data safe?"
everything runs on your computer, with your own claude account. fabrik never stores credentials, and each flow can only use the tools you list. fabrik server only receives run stats and what you choose to share or publish.
"we don't have developers."
installing takes about 20 minutes with a step-by-step guide. the general manager drafts new flows from a sentence; you review before anything runs.
"what if the ai gets it wrong?"
nothing ships without a human yes. ask for changes and the flow redoes only that; write a rule and the flow gets it on every future run.
"what does it cost to run?"
your own claude subscription per person, nothing per task. previas shows the time and usage of every run, so you see payback instead of guessing it.
"we already use chatgpt."
a chat helps one person once. a factory repeats the same quality for the whole team, on a schedule, with a review step and a log.
"is this another platform lock-in?"
flows are plain folders and text files on your disk. stop using fabrik and you keep everything, readable by anyone.
> fabrik request --access
private pilot: a few teams, hands-on setup, honest numbers.
we build your first flow with you, measure what it replaces, and only then talk about price. if it does not pay for itself in the pilot, you owe nothing and keep the flow.
- what we need from you: one repeatable task that eats hours, and one person who can say yes.
- what you get: fabrik installed, your first flow running, a weekly number.
- [PRICE: to be defined after pilot] · no card needed to apply.
we reply within a day · prefer a human? reveal the founder's email.
> fabrik faq
Fabrik Raw Flow: frequently asked questions
What is an AI factory?
An AI factory is a set of repeatable AI workflows (flows), each with a trigger, an input, stations, a human review and a destination, managed from one map. Fabrik Raw Flow is the software that runs them on your own machine.
Which AI does Fabrik Raw Flow use?
Claude, through the official Claude Code program, with each person's own paid Claude account (Pro, Max, Team or Enterprise). Fabrik never stores or shares credentials.
Does it work on Windows?
Yes. Fabrik runs on Mac and Windows as a local app; Windows support is newer and less tested than Mac. For teams, Fabrik Server runs on a server with Docker.
How much does Fabrik Raw Flow cost?
The price will be defined after the private pilot. The running cost is each person's own Claude plan, with nothing charged per task. If it does not pay for itself in the pilot, you owe nothing and keep the flow.
Who builds Fabrik Raw Flow?
Fabrik Raw Flow is built by Gerard Ill, founder of The Big Answer Digital, who uses it every day for his own content and development work. Full profile: thebiganswer.digital/gerard-ill/