CARTIS · Climate Intelligence

Carbon-market answers you can defend.

Climate Intelligence turns standards and methodologies into machine-readable rules, runs them before any AI agent, and keeps a trace of both.

Book a demo

See CARTIS on your own case

Who are we meeting?

CARTIS ontology graph of standards, methodologies, regulations and concepts
CARTIS ontology graph of standards, methodologies, regulations and concepts
Rules first
a deterministic engine runs before any model
6
constrained analysis agents
Versioned
every rule, formula and factor
Graded
a defensibility index on every answer

Why it exists

A carbon project lives or dies on a methodology document. Before spending money, a developer needs to know whether the activity is eligible, what the baseline will credit, which country rules apply and what the credits might be worth. The answers sit in hundreds of versioned PDFs across the CDM, Article 6.4, Verra, Gold Standard, ICVCM and India's CCTS, and a general-purpose chatbot cannot say which version it relied on.

What CARTIS does

CARTIS turns the methodology corpus into machine-readable rules and answers carbon-market and compliance questions in a way that can be defended. A deterministic rule engine and guardrails run before any AI agent, and every answer records the rule versions and evidence it used. It sits at the centre of MARIS and informs every other product.

At work

See CARTIS do the job

Knowledge graph

The methodology corpus, connected

Standards, methodologies, regulations, requirements and concepts become entities in one graph, so a question about eligibility can follow the links a consultant would follow by hand.

  • Entities and relationships, queryable
  • A governed corpus with reliability scores
  • Every document linked to its standard
Ontology graph with type filters and an entity inspector
Ontology graph with type filters and an entity inspector

Agents

Guardrails before generation

An analysis passes through rules and guardrails before the agents see it. If a rule says ineligible, or critical data is missing, the agents never run. What they do say is matched to evidence, and unsupported statements are counted.

  • Methodology, regulatory, financial and integrity agents
  • Rule versions stored with each answer
  • Pure calculations never call a model
Multi-agent intelligence screen with the agent roster and the analysis pipeline
Multi-agent intelligence screen with the agent roster and the analysis pipeline

Standards

Every rulebook in one library

Carbon Registry–India, the CDM, the Paris Agreement Crediting Mechanism, Puro.earth and Gold Standard sit side by side, with their methodologies, tools and documents flowing through one ingestion pipeline.

Standards and methodologies library
Standards and methodologies library

Regulatory

Frameworks, factors and country status

Article 6, CORSIA, the EU ETS and India’s CCTS are tracked with country participation, alongside emission factors and scoring configurations that keep their history.

Regulatory and compliance frameworks with country participation
Regulatory and compliance frameworks with country participation

How it works

An analysis, end to end

Step 01 of 06

Ask

A developer describes the project or asks a question.

Step 02 of 06

Apply the rules

Eligibility rule trees and guardrails run first, with their versions recorded.

Step 03 of 06

Reason

Constrained agents work from the rule engine’s fixed output.

Step 04 of 06

Check the evidence

Each statement is matched to structured references.

Step 05 of 06

Grade

The answer is stored with a defensibility index.

Step 06 of 06

Estimate

Credits and returns are estimated, with a coded diagnostic instead of a silent zero.

Capabilities

Everything in the box

Who it serves:

  • Project developers

    Screen eligibility, estimate credits and plan projects.

  • Validation and verification bodies

    See the same analysis and its trace.

  • Buyers

    Value credits from the buyer's side.

  • Analysts

    Build the corpus and encode the rules.

A governed corpus

Standards and methodologies stored with extracted text, a reliability score and a link to their source.

Rules, not prose

Eligibility rule trees and machine-evaluable baseline formulas for each active methodology, versioned on every change.

Deterministic before generative

No analysis reaches the agents if a rule says ineligible or critical data is missing. Pure calculations never call a model.

Six constrained agents

Agents reason over the rule engine's fixed output, and unsupported statements are counted.

A defensibility index

Each completed analysis stores its rule versions, evidence alignment and a graded defensibility score.

Credit and return estimator

Project and buyer-side estimates that show a coded diagnostic instead of a silent zero.

Principles

How it keeps its integrity

  • Rules run before reasoning.
  • Models draft, people activate.
  • Every change has a version.
  • Claims need evidence.
  • Fail loudly: a formula that cannot be evaluated says so.

Clear scope

What it isn’t

Not a rating agency
It grades the defensibility of its own answers, not projects or credits for sale.
Not a registry
Issuance, transfer and retirement belong to AMMONIS.
Not legal advice
It reports what the encoded rules say.

Built on

What CARTIS stands on

The platform underneath, and what the work depends on outside it.

The platform

  • One platform core

    The same evidence architecture the rest of the suite runs on: identity, roles, attribution and the audit trail are not rebuilt per product.

  • A deterministic rule engine

    Eligibility rule trees and machine-evaluable baseline formulas run before any agent does.

  • An ontology over the corpus

    Standards and methodologies are held as linked, versioned rules rather than as documents to search.

  • Constrained agents

    Agents reason over the rule engine's fixed output, and every answer records the rule versions and evidence behind it.

The rules it holds

Registries and programmes

  • CDM
  • Paris Agreement (PACM)
  • India CCTS
  • Carbon Registry–India
  • Verra
  • Gold Standard
  • Global Carbon Council
  • Puro.earth
  • Rainbow

International standards

  • ISO 14000 series
  • GHG Protocol
  • ICVCM
  • CORSIA
  • CRCF
  • ICROA
  • FSC
  • PEFC

Emission factors

  • IPCC
  • DEFRA
  • National grid factors
  • Published allometry

Reporting and compliance

  • BRSR
  • SBTi
  • CDP

Representative, not the whole corpus.

Questions

Asked often

How is this different from a chatbot?+

A chatbot cannot tell you which version of which rule it relied on. CARTIS runs encoded rules first and stores the versions, the evidence and a defensibility grade with every answer.

Does it rate credits?+

No. It grades the defensibility of its own answers. It does not rate projects or credits for sale.

Which rulebooks does it cover?+

The CDM, the Article 6.4 mechanism, Verra, Gold Standard, ICVCM and India’s CCTS, each versioned as they change.

Is it legal advice?+

No. It reports what the encoded rules say.

See CARTIS on your own data

Tell us about your programme and we’ll walk you through it.

Book a demo