ClauseSpace

Your footprints into the data space, and your terms in plain words.

Model calls: not configured (no API key)
  1. 1Your data
  2. 2A few questions
  3. 3Your wishes
  4. 4Publish & live requests

What is a data space, and why does it matter?

Large manufacturers now need data from their suppliers. A car maker, for example, has to report the carbon footprint of everything it buys, so it asks each supplier for the footprint of each part. E-mailing spreadsheets doesn't work at that scale, and once a file is sent you lose control over it.

In a data space every company runs a connectorA piece of software that guards your data. It publishes offers, checks who is asking, and only hands data over after the other company's connector has agreed to your rules.. You keep your data, publish an offer with rules ("only my customer", "delete it by 2027"), and another company's connector must accept those rules before it gets anything. The automotive industry's data space, Catena-X, works this way, and more industries are following.

For a small supplier the hard part is getting in: the footprint format has about 90 fields, and the rules are written in a technical policy language.

What ClauseSpace does

It takes a small supplier from the spreadsheet they already have to a working offer in a data space, in four steps:

  1. Your data: an AI (NVIDIA Nemotron, in the EU) matches your columns to the official footprint format; code checks every match and converts the units.
  2. A few questions: only what your sheet doesn't say, with explanations and suggestions.
  3. Your wishes: you say in plain words who may use your data. ClauseSpace turns that into rules and tells you honestly which wishes the connector can enforce and which it can't.
  4. Live requests: your offer goes into a real connector, and you watch other companies ask for it: your customer gets in, the others are refused.

Takes about 3 minutes with the sample file. All companies in the demo are fictional.

Step 1 of 4

Your data

What happens: ClauseSpace reads your footprint spreadsheet and matches each column to a field of the official format (PCF data model, version 9.0.0).

What you do: upload your file, or try the sample: the footprints of four metal parts, as a consultant might send them.

Your rows stay on our server. The AI (NVIDIA Nemotron on Nebius Token Factory, EU region) sees only column names, detected types and units, and made-up example values.

Try the sample

Feldhagen Präzisionsteile GmbH (fictional): 4 stamped metal parts, 25 columns with the usual mess: grams and kilograms, values per kilogram instead of per piece, internal columns that must not be shared.

Upload your own file

CSV or Excel, one row per part. Up to 1 MB, 50 columns, 1,000 rows.
Want a harder test? The same parts as a German export (German headers, decimal commas, dates as DD.MM.YYYY):
Stats for nerds · 0 model calls, 0 tokens, 0 s, $0.0000 this session

Model calls in this session

Routing rule: anything derived from your data goes only to models in eu-north1 (here nvidia/Nemotron-3_5-Lightning, eu-north1); code checks this before any network call. Prompts are never logged. The model sees column names and made-up examples, never your rows.

No model calls yet.

Honesty test

Last full run 2026-10-30 18:06 UTC: passed (8/8 tests). It proves that every rule shown as "checked by the connector" refuses a request that breaks it, and every "binding" rule never changes the outcome.

Token limiter

4000

Worst case per call on the default model: .