Start building

Executable knowledge infrastructure

Put executable knowledge to work in your apps and AI agents.

Executable models of laws, standards and textbooks give your applications reproducible results with grounds you can verify.

Send facts and a question to a versioned knowledge model. Get the result, the rules it applied and the source edition it used.

Start building Try the example
EUR-Lex · 261/2004example edition · 08.2026

Passenger right to compensation

case2,300 km · cancelled 5 days ahead · no re-routing offered

View source text

5.1(c)Passengers informed of a cancellation less than seven days before departure are entitled to compensation, unless offered re-routing close to the original schedule or the carrier proves extraordinary circumstances.

7.1(b)For all intra-Community flights of more than 1,500 km and for all other flights between 1,500 and 3,500 km the amount is EUR 400.

Model · eu.transport.air_passenger_rights@0.1.0

Arxo executes the canon.

Facts → Articles → Result

Change the facts. Check the result.

Change the distance or the notice period and run the flight case again. The other tabs run inheritance rules, a chess game and literature on the same engine.

Your question

A 2,300 km flight was cancelled 5 days ahead. What compensation is due?

Chain of reasoning4 grounds
  1. Art. 5(1)(c)(iii)Notice came less than seven days ahead and no re-routing was offered, so the right to compensation arises.
  2. Art. 7(1)(b)The flight is longer than 1,500 km and not longer than 3,500 km.
  3. Art. 7 tableThe only applicable amount for that range is EUR 400.
  4. Art. 5(3)The carrier does not invoke extraordinary circumstances, so the exemption does not apply.
Result€400

The carrier’s duty arose. This is the only conclusion consistent with the established facts.

sample result · fixed facts and edition
  • Source and editionThe text and the version the rules come from.
  • GroundsThe provisions behind each step from the facts to the result.
  • Clear answer statesThe outcome, the facts still needed, and any decision assigned to a person.
  • Same result every timeThe same facts and the same edition give the same result.
Watch the walkthrough · 2:23 Ten business days from 6 March 2026, on Lens (document in Russian) How to read an answer

Reuse knowledge with its grounds.

A source can guide many applications. Arxo makes its interpretation a versioned model that your software and AI agents can use as a dependency.

The model defines how stated facts lead to a conclusion. Its grounds let you inspect the rules applied, the source edition used and the questions still open. When the source or the facts change, you have a model to revisit and a result to examine.

  1. Source
  2. Explicit model
  3. Facts and question
  4. Result with grounds

Reuse the model. Preserve the conditions. Give the next person something they can check and build on.

Explore the computational model Read the Arxo Manifesto

Update the rules without retraining your AI.

A canon is a versioned package of rules — a law, a standard, a reference text — linked to its source text. Add it to your application or agent as a dependency instead of training it into model weights. When the source changes, you update one package.

  • Knowledge as a dependencyPin a canon version like any other dependency. The AI model that asks the question stays as it is.How a model is versioned
  • Every edition on recordAn edition applies to a case on a given date, and earlier results replay on the edition they used.See an executable edition of the law
  • One layer, many modelsYour product, your agents and public services ask the same canon and receive the same result with the same grounds.

Choose how Arxo fits your stack.

The same model reaches your application, your AI agent or your own codebase.

MCP

For an agent. Query a named model version and read the answer with its grounds. The public endpoint receives the question and the facts; do not send personal data.

Connect an AI agent

Generated code

For your own codebase. The module returns the answer and its status. The rules, passages and edition come back through the SDK and MCP.

See target languages

Have your own rule, standard or policy? Turn it into a knowledge model →

For legal and policy teams: impact of amendments, unintended outcomes and the map of a dispute →

One mechanism for many areas of knowledge.

Arxo can model any area where an authoritative text sets the rules: laws and treaties, technical standards and protocols, clinical scores, textbooks, the rules of games. Law comes first; the same mechanism runs the rest.

Laws
EU Regulation 261 · GDPR · Code civil · BGB · Codes of Kazakhstan
Treaties and trade
UN Charter · CISG · Incoterms
Technical standards
IEC 60364 · NEC · ISO 13616 IBAN · GUM measurement uncertainty
Web and security
CSS cascade · HTTP semantics · OAuth 2 access tokens · PKIX certificate paths · Open-source licences
Clinical scores
CHA₂DS₂-VASc · HAS-BLED · PHQ-9 · GAD-7Models of the named instruments. A result is not a diagnosis or a treatment.
Science and textbooks
OpenStax kinematics · OpenStax integrals · Newton’s Principia
Games and sport
FIDE Laws of Chess · IFAB Laws of the Game
Religion, history, literature
Zakat · Faraid inheritance shares · Magna Carta 1215 · Dante’s Inferno
Browse the law and doctrine canons Search every area through MCP The ideas behind Arxo

What will you put to work?

Bring a rule, formula, or requirement you already use. Start with one calculation or check.

Discuss your use caseModel your own source