Skein runs on your machine. Your files stay in your building, no account is created, and nothing is sent anywhere. That is not a setting you switch on, it is the only way it works.
| Question | Answer |
|---|---|
| Where does my data go? | Nowhere. It is read on your machine and written to one file on your machine. There is no upload step because there is no server to upload to. |
| Can anyone else see it? | No. Each business is a separate database file, not a row with an owner column, so there is no query that could return somebody else's records even by mistake. |
| Does it call an AI service? | No. There is a test in the build that walks every import and fails if any module capable of opening a socket appears. |
| What does it cost to try? | Nothing, and there is no sign-up. Delete the folder and it is gone. |
Almost every Mac does. Open Terminal, or Command Prompt on Windows, and type this. Any version of 3.11 or later is fine.
python3 --version
If it is missing, install it from python.org. Nothing else is needed: Skein carries its own database drivers, so there is no second install to chase.
Double click Start Skein in the folder. On a Mac that is
Start Skein.command, on Windows Start Skein.bat. A
browser tab opens on your own machine.
The first time, macOS may say it is from an unidentified developer. Right click the file and choose Open, and it will offer to run it anyway.
One box, one name. That creates your graph, which is a single file on your disk. Two businesses means two files, and they cannot see each other.
Go to Sources. Three kinds of thing go in:
A folder of documents. Give it a path. It reads Word, PDF, text and markdown, and tells you what it skipped and why rather than failing silently.
A spreadsheet. Export to CSV and point at it. You choose which column names the record.
Your database. Six boxes, not a connection string: kind, host, port, database, user, password. Read-only is enough, and read-only is what we recommend. Skein shows you what it intends to do with every table before it reads a single row, and you can change or refuse any of it.
Pick any client from the list. You get what you would have written before a call: current terms and what they replaced, figures on record, what is open and late, who is involved, and a section saying plainly what it does not know.
That last section is the one to judge us on.
It re-reads only what changed. On a live PostgreSQL of 15,699,911 rows the first import took 2.2 minutes and a re-index with nothing changed took 0.12 seconds, which is what makes an hourly refresh affordable rather than theoretical. Large fact tables are summarised inside your database and never travel: fifteen million of those rows cost three seconds.
The full report goes table by table and keeps the three defects we hit while measuring.
Where did it feel unprofessional or half finished?
Is the brief the page you would actually want before a call?
Did anything make you distrust the numbers?
Where did you get stuck, and did the screen tell you what to do?
Onboarding has been run properly by one person, a handful of times. That is the part most likely to break in front of you, and the part we most want you to try cold.
No SOC 2 and no HIPAA agreement. For an install where the data never leaves your own machine the audit surface is yours rather than ours, which is an argument and not a certificate.
Skein by RyanAI LLC · Missouri · Stuck on any step, say which number and what the screen said: ry@ryanai.io