For software vendors

High accuracy emission factor matching, inside your product

Your users upload a procurement file to you. You send the lines to one endpoint and get back the factor, the library, the boundary, the emissions and the reasoning for each one, all drawn from open data anyone can check. The hard part of Scope 3, without a data team of your own.

In production in a few days. Not a quarter.
POST/v1/matchbatch, 5,000 lines
{"description": "ACIER BETON HA B500B 12MM""quantity": 3000"unit": "kg""country": "FR""line_id": "client-42"}
Response
{"activity": "Reinforcing steel, hot rolled""library": "ADEME 23.6""region": "FR""boundary": "cradle_to_gate""emissions_kgco2e": 6990"confidence": "high""rejected": ["Steel, low alloy", "Steel section, hot rolled"]"reasoning": "B500B is a reinforcing grade, not a structural section."}

One call per batch. The factor, the library, the boundary, the emissions and the reasoning come back on every line.

96%of matched lines usable as returned, on the open benchmarkread the benchmark
4,673,506emission factor values synced into your databasemore comingsearch the open data
101emission factor libraries, normalised into one schemamore comingbrowse libraries
2 htime to first result in your MCP

What it changes for your product and your P&L

A footprint in the first session

Your user uploads the file they already have and sees a number, instead of being asked to fill in a template first.

Try it

Quality that makes your product scale

Matching the easy lines is not a product. Your users only save time if the hard ones come back right too, and that accuracy is the number we publish and keep measuring.

Your engineers on your roadmap

No ingestion pipeline, no factor curation, no unit vocabulary to maintain. One endpoint and a schema that does not move.

20x cheaper than building it

€100k to €300k a year of engineers, plus the language model tokens they run through, against matching from €7,500 a year with the tokens included. The open data itself is free.

What you save

The data stays yours

Your account, your keys, your region and your data-processing terms, with usage and logs on your side.

Library management off your plate

Publishers revise, retire and renumber. That lands here and reaches you as the same schema it always was.

What we normalise

Three ways to plug in

Search, match or sync, from your own product

Make or buy

20x cheaper than building it

Building it means engineers, language model tokens on every line and a fresh evaluation every time the models move, for as long as you ship it. Here all three are in the price, and the open data is free.

20xless than building it and keeping it running

  • Building it yourself€100k to €300k a year
  • Matchingfrom €7,500 a year, tokens included
  • Search and open data€0

€0.050 a matched line, down to €0.015 at volume

See pricing, starting at €0

What you stop paying for

  • One to three engineers

    Ingesting every library, following every revision and keeping the matching accurate is a team's job, for as long as you ship it.

  • Language model tokens

    Reasoning over every line costs tokens, and so does every re-run and every evaluation. Ours are in the price.

  • Lines you cannot use

    Credits go only on high and acceptable matches, and one material error refunds the whole file.

  • Keeping up as models move

    We re-test the matching whenever the models change, so a better one reaches you without a release on your side.

See everything it takes to build
What it takesBuild itUse open-climate.ai
Ingesting and normalising every public libraryOpen dataYour teamIncluded
Following every publisher revisionOpen dataYour teamIncluded
One boundary, unit and warming vocabularyOpen dataYour teamIncluded
The matching itself, at a defensible accuracySee what features you needYour teamIncluded
Language model tokens for every line, re-run and evaluationYour billIncluded
A fresh evaluation every time the models moveYour teamIncluded
The reasoning trail an auditor will ask forSee what features you needYour teamIncluded

Time to first result

Hours over MCP. Days over REST.

  1. Day 02 hoursOver MCP

    A key, the MCP server pointed at your assistant, and a file matched end to end.

  2. Week 12 to 3 daysOver REST

    The REST endpoint wired into your upload flow, with the confidence bucket driving your review screen.

  3. Week 2your callYour own factors

    Your own factors matched beside the open ones, and your identifiers returned alongside ours.

What people say

In a couple of hours, open-climate.ai's MCP was integrated into Carbon+Alt+Delete.
HanspeterCTO, Carbon+Alt+DeleteLeading GHG accounting tool · Belgium

Put it in front of your customers this quarter

Tell us what your users upload and where the result has to land. We will work out the shortest path to it.

Choose a filexlsx or csv, any shape. No template, no cleanup week.
  • EU hosting
  • Never used to train language models
  • Confidentiality guaranteed