Regixo docs
The whole journey, end to end

The whole journey

Start with the data catalog — map your data, connect your sources, keep it fresh. That is a complete product on its own, and it is where everyone begins. If your company is in the EU and must keep a GDPR RoPA or DORA record, an optional compliance module continues from the same map. Each stage below links to a walkthrough that walks it, screen by screen.

Prefer to be walked through it? The Walkthroughs follow real people through a live session: Your first session (engineer) · The compliance team · The hand-off round-trip · An afternoon with the agent · A change & a re-sign. This page is the map; those are the lived experiences.
Two paths through this page You run things (engineer): start at step 1 and work down. Someone forwarded you a record (compliance): the steps that are yours start at step 7 — jump to Understand the record → and skip the install steps entirely. You install nothing.
Two parts The data catalog (steps 1–4) is free and for everyone — the map is the product. EU compliance (the optional continuation below) is only for companies that must keep a GDPR RoPA or DORA record. Skip it entirely if you don’t.
Say it, or run it

Every step you run comes two ways: the sentence you say to your own coding agent — Claude Code, Cursor, Codex, already open in this project — and the command it runs.

Your agent runs the same regixo command in the same shell. No second path, nothing extra to install. The switch at the top right chooses which one leads, on every page.

Steps 7–9 belong to your compliance team, and signing carries no sentence at all — that one is a person's.

Running headless, in CI, or from a script None of this needs a person at the keyboard. Every command takes --json; setup (start, add) takes --yes, --non-interactive and --intent. A missing required input comes back as a coded error that names the flag to supply — never a blocking prompt.

1 · Get your data mapped

You need Node ≥ 22.18 — nothing else. One command scans your sources’ structure into a local catalog and drafts your record. It asks at most two questions.

say

“Set Regixo up in this project — scan my sources and build the data map.”

Show the commandHide the commandShow the sentenceHide the sentence
run
$ npx regixo start
then

Your agent runs the scan — it reads Regixo’s own manifest first, so it is not guessing at commands, then reads names and types only, never a row value. Whether you are a regulated financial firm is yours to declare: Regixo cannot infer it, and it decides whether a DORA register is drafted.

Check it worked: a dataset count, a count of columns that look like personal data, and a DRAFT record. Then read regixo statuscoverage line: a source it could not reach is missing from the map, the counts are quietly too low, and the coverage line names it.

Show what it prints in the terminalHide the terminal outputShow what your agent reportsHide what your agent reports
example output
Regixo — free to install and use. It maps your data and drafts your compliance record, right here. The scan stays on this machine.

▸ Detecting data sources…
   Postgres        DATABASE_URL       reachable
  ? Connect & scan this source?                     [Y/n] y
  ? Is your organisation a regulated financial entity?  [y/N] n
▸ scanning metadata…  63 tables · 55 with personal data
 your data map is ready           →  regixo open
 draft record generated           →  RoPA_DRAFT.html

Full detail, the two questions, and every state: Install & first run →

2 · Connect your sources

If nothing was auto-detected, connect a source by hand. Regixo stores only the name of the env var that holds your secret — never the secret itself.

say

“Add my Postgres database to Regixo.”

Your agent fills in everything except the token — you put that in .env yourself.

Show the commandHide the commandShow the sentenceHide the sentence
run
$ regixo add postgres --ref DATABASE_URL --label "App DB"

Every connector (Postgres, MySQL, SQL Server, Redshift, Snowflake, BigQuery, Stripe, dbt, CSV), plus your own connector for anything else (regixo connectors new / regixo add script), the real connection strings, and how to reach a locked-down database: Connect your sources →. Four have not been run against a live account yet: Snowflake, BigQuery, Stripe, and Redshift — which rides the verified Postgres driver.

3 · See the map

Open the catalog in your browser. Search it, filter to the personal data, see why each column was flagged and where data flows.

run
$ regixo open

Run this one yourself, not through your agent — it keeps running until you press Ctrl+C, so give it a terminal of its own.

Give it a terminal of its own

regixo open is not like the other commands: it does not finish. It keeps running until you press Ctrl+C (hold the Control key and press C) — and while it runs, that terminal is busy. So open a second terminal window, run it there, and leave it running. Do everything else in your first one.

If you work with a coding agent, run this one yourself. Don't ask your agent to open the portal — not even if it offers to run things in the background. A portal your agent starts belongs to its session: you can't stop it, you can't restart it, and it disappears the moment the agent finishes or moves on. Your page then looks fine until you click something — and that click fails with "localhost refused to connect", with nothing on screen to connect it to a command your agent ran a minute ago. It looks like Regixo broke; nothing broke.

Leave it open while you work. The portal re-reads your catalog every time a page loads, so a re-scan, a new description or a corrected flag — yours, or your agent's — shows up as soon as you refresh. No restart needed.

Settings are the exception. regixo.yml and your environment variables are read once, when it starts. Change a setting there and the running portal won't notice: stop it (Ctrl+C) and run it again.

regixo serve-claim behaves the same way, and so does regixo mcp when your AI tool launches it. Every other command runs and hands the terminal straight back.

What the scanner reads vs refuses, and how the classifier decides: Read the map →

— or in the portal: regixo open is the web portal — browse and correct the map by pointing and clicking, no commands. Tour it: The free portal →

The two agent surfaces — don’t confuse them

An agent works with Regixo two ways, and they are separate. You have been using the first one since step 1.

1 · Operating Regixo — you are already doing it

Say the sentence to the coding agent that is open in this project, and it runs the regixo command for you. There is nothing to install and nothing to register. It is the same command in the same shell — that is why every step on this page shows you both.

To hand over the whole job at once, give it the playbook. It teaches the setup order, the guardrails, and where the agent must stop and hand back to you:

say

“Read Regixo’s operating playbook, then set it up for this project.”

Show the commandHide the commandShow the sentenceHide the sentence
run
$ regixo skill

2 · Reading the catalog — a real, one-time setup

The other surface is an assistant that reads your catalog and answers questions about it — "which columns look like email addresses?" — over MCP, the protocol Claude Desktop, Claude Code and Cursor already speak. It is read-only by construction: there is no write tool at all, so it can never scan, correct a flag, or change the record. Unlike surface 1, it takes a registration:

$ regixo mcp --print-config

It prints a JSON block you paste into your assistant’s MCP settings, once.

Why the difference matters An assistant reading the catalog over MCP cannot correct a personal-data flag — that surface has no write tool. Your coding agent can, by running regixo classify set, because that is an ordinary command and a flag is a mechanical fact. What neither may ever do is make a legal call — confirming a purpose, a lawful basis or a retention period, or signing the record. Those carry no sentence anywhere in these docs, on purpose.

Both surfaces — the eight read tools, the tasks-to-commands table, and what a session looks like: Use an AI agent →

What you can add — and what you can skip optional

There are four things a person can add on top of the scan: correct a personal-data flag, draw a data flow the scanner cannot see, describe a table, and define a business term. None of them is required. Your map is done, and so is your draft record — the hand-off below does not wait on any of this.

They are also not equivalent. Two of them change your record — your categories of personal data come from the flags, and a cross-system flow is suggested as a recipient under Art. 30(1)(c). Two of them change nothing in it at all: a description and a glossary term never reach the compliance engine. That is not a reason to skip them — a description is the one thing you write that everyone else reads, and it is how a person who does not know your schema works out what a table is for. It is a reason to know which is which before you spend an afternoon on one.

Which ones are worth your time, given what you came here to do: What you can add →

And when someone who doesn’t run Regixo asks to see the estate: regixo share writes one self-contained file — a briefing for your compliance team, with the full map underneath as evidence. Metadata only; it opens in any browser and reaches nothing on the internet when opened.

4 · Keep your map current

Your schema changes, so re-scan on a schedule. regixo watch refreshes the map and logs exactly what changed — added, removed, reclassified — so your catalog never drifts. Run it from CI and you are done.

say

“Schedule Regixo to re-scan my data automatically and tell me when something changes.”

Show the commandHide the commandShow the sentenceHide the sentence
run
$ regixo watch --ci

The CI recipe and the change log: Keep it current →

That is the whole data catalog — a complete product on its own. Everything below is optional.


Optional · EU compliance — everything below The steps below are only for EU companies that must keep a GDPR Article 30 RoPA or a DORA register. Regixo already drafted that record from the same map — so if you don’t need it, you are done above. The catalog stands on its own.

5 · The draft record

The same first scan already drafted your GDPR Article 30 record. It is honest about what it can’t know: the data categories are filled in, but the blanks are marked needs you.

Three of those blanks are yours — and nobody else can close them. Your compliance team does not know whether your database is encrypted, which outside services get a copy of the rows, or whose data is in which table. Those three questions live in “Your part” at the top of the Record page in the portal: tick the boxes, press save, done in about a minute. Leave them blank and they reach your compliance team as gaps — and the only way to close one is an email back to you.

Everything else — purpose, lawful basis, retention — is a legal judgement. That part is genuinely not yours.

The line Regixo won’t cross Regixo attests mechanical facts only — purpose, lawful basis and retention are confirmed and signed by a human, never by the tool or an AI agent. This is not legal advice.

What the record contains, field by field: Understand the record →

6 · Hand it off

Forward the draft to your compliance team. Before anything leaves, Regixo shows you exactly what will be sent — metadata only — and asks you to confirm.

say

“Forward my Regixo draft record to our compliance team. Show me what would be uploaded before you send it.”

Show the commandHide the commandShow the sentenceHide the sentence
run
$ regixo invite
then

This one asks before it sends. It previews exactly what leaves the machine and waits for a yes. What goes is the full name of each table that looks like it holds personal data — source, schema, table — plus the drafted record. What does not go: your column names, your column types, and never a row value.

Check it worked: you get a DRAFT PDF and a claim link your compliance team opens with no account. Your agent must show you that preview and get your word first — the playbook tells it never to assume consent. If it sent without asking, say so.

Show what it prints in the terminalHide the terminal outputShow what your agent reportsHide what your agent reports
example output
About to upload the draft snapshot to https://app.regixo.com (this is the DEFAULT — it leaves your machine).
  what leaves the machine: The full name of the 4 tables that look like they hold personal
  data — source, schema and table (for example “app-db/public/customers”) — and for each one
  the category label the classifier derived from the column names (“Email”, “Name”). The
  drafted record: activity titles and Regixo’s suggested legal fields. Everything else is a
  count, not a name: 2 sources, 41 datasets, 118 personal-data columns. It is metadata only:
  no column names, no column types, no row values, ever. Owner emails are hashed before sending.
Upload it? [Y/n] y
 RoPA_DRAFT.pdf written — forward it to your compliance team
 claim link: https://app.regixo.com/claim/clm_7Q…

Or do it from the portal — no terminal

If the portal is open (regixo open), the record page does the whole hand-off. Open Record and use the Hand it to your compliance team card:

Once it is made, the card shows the live link with a Copy link button and a Download the draft PDF button. If the upload did not happen — you declined it, or the portal refused it — the card says so plainly and offers no link to copy: a link nobody can open is never handed to you as though it worked.

Read it before you send it — regixo serve-claim

The draft is built from your table and column names, and you are about to put it in front of your compliance team. It is worth reading first — a table name can say more than you expect, and this is the moment to notice. regixo serve-claim opens the record they will read, on the hosted portal’s own page, from your machine, with nothing sent:

run
$ regixo invite --no-upload   # draft it — send nothing
$ regixo serve-claim          # read what they'll read

Run this one yourself, not through your agent — it keeps running until you press Ctrl+C, so give it a terminal of its own.

then

The order is the point. regixo invite on its own uploads (it asks you first, but a preview you run after sending is not a preview). --no-upload builds the very same draft and sends nothing — so you read it, and then decide.

Check it worked: a browser opens on the record. Read the Your privacy line: it tells your compliance team, in their own words, that only names and types left your machine and no row values, ever. Read and expand anything — but nothing can be saved, signed or sent from here, and clicking a control says so. When it reads right, run regixo invite to send it.

Show what it prints in the terminalHide the terminal outputShow what your agent reportsHide what your agent reports
example output
▸ starting the claim preview on port 4320… it will keep running here — press Ctrl+C to stop (use a new terminal for other commands).
 Claim preview serving at http://localhost:4320/claim/clm_7Q…   (the record they will read)
  opening your browser…
  ⏳ this terminal is now running the server — press Ctrl+C to stop.
  need other commands? open a new terminal.
  ↳ read and expand anything. Nothing can be saved, signed or sent from here.
  ↳ the record is what they will read; the sign-in and unlock steps belong to the hosted portal.
  ↳ reads right? send it with `regixo invite`.
What the preview can’t show you Only the record is theirs. The rest of the page is local: the make-it-official card shows your machine’s state — no licence — rather than the plan and price the hosted portal will offer them, and a DORA register does not appear at all (the snapshot carries only its table count). Read it for the record, not for the sales page.

What the hand-off sends, and what your compliance team receives: Claim & review →

The hand-off seam — from both sides

The engineer finishes here. You answer your own three Art. 30 fields — security measures, recipients, data subjects — then forward the draft. The page that walks both, without leaving your editor, is Answer your part & forward →. Two states worth telling apart: an invitation you made (a metadata-only snapshot was uploaded and you get a claim link) versus a draft you only shared to read (regixo serve-claim, or --no-upload — nothing left the machine, and no link is handed to you).

The compliance team picks it up at step 7. They open the claim link with no account, read the DRAFT, make the legal calls, and unlock — and the price is on the page, worked out from two answers, never “ask us”.

7 · Claim & fill the legal calls compliance team

Your compliance team opens the claim link, signs in with a one-time email link, and the record opens. They make the calls only a human can: a lawful basis for each activity, retention, special-category grounds, and any DORA contract data. Regixo suggests; a person decides and confirms.

Every legal field, the real Art. 6/9/10 options, and how to choose: Fill the RoPA → · financial firms: The DORA register →

— in the portal: this all happens in the browser, no install. The full screen-by-screen tour: The compliance portal →

8 · Unlock & sign compliance team

When the calls are made, unlock to turn the draft into a signed, defensible official record. The price is published and worked out from two answers.

Pricing €6,000 (RoPA) · €12,000 (RoPA + DORA) · from €18,000 (enterprise), per year. Never “ask us”. Regixo is pre-1.0 and has not launched — nothing is purchasable today; the prices are published ahead of it.

What signing does, the seal, and re-signing when things change: Unlock, sign & maintain →

— in the portal: Unlock & pay then Sign & seal are portal screens. See them: The compliance portal →

9 · Re-sign when things change compliance team

The same regixo watch that keeps your map fresh also flags any signed activity whose core fields moved — so your compliance team can re-sign and the record stays defensible. The team’s fills also come home to the engineer’s machine.

Re-signing, the change log, and the fills coming home: Unlock, sign & maintain →