GreenCalculus

The carbon API that hands back the source with the number.

16,000+ sourced greenhouse-gas emission factors and audit-traced calculations — every value returned with the publisher, the exact cell it was read from, and the data version it was read at. So your users get a figure an auditor can follow, not a guess.

Get a free key · SDKs · MCP · Open data · Benchmark · greencalculus.com

The corpus is open to read — no API key:

curl "https://api.greencalculus.com/v1/factors?key_prefix=grid.gbr.electricity.location_based&limit=1"

{
  "key":   "grid.gbr.electricity.location_based",
  "factor": { "value": 0.13096, "unit": "kg CO2e per kWh", "gwp_set": "AR5_100" },
  "source": { "id": "DEFRA_2026", "cell_ref": "'UK electricity'!E25", "retrieved": "2026-06-18" },
  "licence": { "name": "Open Government Licence v3.0", "redistributable": true },
  "citation": { "proof_url": "https://verify.greencalculus.com/grid.gbr.electricity.location_based@2026.188" }
}

That proof_url is a permanent page showing the publisher, the document, the exact cell and whether the value may be republished — a link your own reader can check.

Searching the corpus is keyless too:

import { GreenCalculus } from "greencalculus";      // npm install greencalculus

const gc = new GreenCalculus();                     // no key
const { factors } = await gc.search("diesel litre");
console.log(factors[0].key, factors[0].factor.value);   // fuels.can.diesel.litre 2.68901

A free key (no card) adds the seven calculation engines and as_of= version pinning.

Seven calculation engines — GHG Protocol activity, electricity (location- & market-based), freight, business travel, spend-based EEIO, embodied carbon (EN 15978), and PCAF financed emissions. Each returns the full working and a deterministic receipt hash, never just a total. as_of= pins any factor to a past data version, so a figure re-runs identically in an audit.

{ "mcpServers": { "greencalculus": {
  "url": "https://mcp.greencalculus.com",
  "headers": { "Authorization": "Bearer gc_live_..." } } } }

Discovery and the keyless tools work without a key. explain_absence answers why a factor you expected isn't there — usually the real question behind "do you have X?".

Ask a language model for an emission factor with no tools and it will usually give you one. greencalculus-benchmark measures how often that number is right — and how often the model names the wrong source for it. 467 questions, 45 sections, 75 publishers, generated deterministically against sourced ground truth.

Five models, 467 questions, no tools:

The more a model answers, the less each answer is worth — and the last column sorts with talkativeness. A model that cites DEFRA for an EPA figure is more dangerous than one that cites nothing, because the citation is what makes a reader stop checking.

Connect GreenCalculus and it goes away. 90 questions, paired, same model with and without two keyless lookup tools:

About +61 points for both, at 1.0–1.6 tool calls per question. The model isn't handed the answer — it searches, picks the factor and reads the value, so this measures the integration, not a rigged prompt. Full write-up, caveats, and the log of the scorer's own six bugs.

greencalculus-open-data publishes the coverage map — 16,673 factors from 137 publishers, each row with its key, section, unit, gas, GHG Protocol scope, publisher, licence, and whether it may be republished. Generated from the live API, nothing hand-maintained.

• 15,347 factors / 75 sources carry a licence that permits republication — OGL, CC BY 4.0, Etalab, US public domain, Eurostat reuse.
• 1,326 factors / 62 sources do not, and the reason is the publisher's own: paid standards, NonCommercial or NoDerivatives terms, viral share-alike, or no grant found.

No factor values — it's our metadata about what the corpus covers, so it's publishable regardless of upstream terms and you can use it freely to answer "is there a factor for X?" without worrying about anyone's licence. Values come from the API, keyless. LICENCES.md carries every source, its licence, and the attribution line that licence requires — plus every excluded source with the publisher's stated reason, verbatim.

One versioned, date-anchored factor store — not scattered hardcoded constants. The same factor feeds the public calculators, the API, the SDKs and the MCP server, so the maths can never disagree with itself.

• Versioned — factor sets are date-stamped (2026.188); a value change is a tracked, auditable event
• Validated — automated parity checks gate every change, so no surface silently drifts from the canonical value
• Traceable — every value carries its publisher, cell reference, licence, uncertainty and GWP basis
• Fail-soft — an unrecognised key degrades visibly, never to a wrong-but-plausible number

Free public tools at greencalculus.com, organised as families rather than one-offs — each folds regional and per-asset variants into a single tool:

• Scope 1 — stationary & mobile combustion, refrigerants, process emissions
• Scope 2 — location- & market-based electricity, steam, heat, cooling
• Scope 3 — spend-based (MRIO/EEIO), categories 1–15, transport & distribution
• Financed emissions (PCAF) — all 7 Part A asset classes, plus facilitated, insurance-associated, and data-quality scoring
• Land sector (FLAG) — land-use change, land management, and removals as separate lines
• Target setting (SBTi) — near-term Absolute Contraction trajectories and net-zero pathways

Aligned to GHG Protocol Corporate & Scope 3 · LSR (2026) · IPCC AR6 GWP-100 · ISO 14064-1 · CSRD/ESRS E1 · SBTi · PCAF.

Every factor response ships a ready-made citation line:

UK grid electricity — location-based (generation). UK Government GHG Conversion Factors 2026 — Department for Energy Security and Net Zero (DESNZ), cell 'UK electricity'!E25, retrieved 2026-06-18. via GreenCalculus data version 2026.188.

For a calculator or methodology page:

GreenCalculus (2026). [Title]. Version [x]. GreenCalculus.com. https://greencalculus.com/... (accessed YYYY-MM-DD).