The AMX Platform

AMX is a symbolic verification engine — not an LLM, not a wrapper, not a guardrail. It is a deterministic reasoning system with zero neural components that independently re-derives AI answers from source evidence.

Three-Layer Stack:

• BigLedger — Enterprise business software foundation (5,000+ businesses, 100+ applets)
• AMX — Symbolic verification engine (the core reasoning layer)
• LLM Bridge — Connects to external LLMs (GPT-4, Claude, Llama, Mistral) for natural language tasks; AMX verifies all outputs

Products

AMX CLI

The developer command-line interface to the AMX platform.

• amx build — build and run verifications against Knowledge Capsules from the command line
• amx cps — full capsule lifecycle management: init, build, test, validate, sign, and publish capsules to any Maven-compatible repository
• amx mcp serve — expose AMX verification as MCP tools (amx_reason, amx_verify, amx_replay, amx_explain) for integration with any MCP-compatible agent or IDE
• amx agent — configure, deploy, and publish Coworker agent cards to your infrastructure or AMX Central
• Proof certificate inspection — view, validate, replay, and export XFPC proof certificates
• Pipeline orchestration — chain verifications, connect to LLM Bridge, and automate workflows
• Environment management — switch between development, staging, and production configurations
• Native on all platforms — runs natively on macOS, Windows, Linux, and JVM environments

Enterprise Agentic Twin

A business digital twin platform for hosting, simulating, and running your organization as a verified virtual environment. Design it, simulate it, then deploy it — with every decision verified.

• Digital twin hosting — run a live digital replica of your organization: departments, workflows, decision chains, and resource flows
• Design → Simulate → Deploy — model your organization digitally, test business scenarios with verified outcomes, then deploy with confidence. Model policy changes, staffing decisions, pricing strategies, and process modifications before committing real resources.
• Coworker agents — role-based verified AI twins (Finance Coworker, Compliance Coworker, Operations Coworker, Risk Coworker) that each own specific business functions. Each Coworker uses LLMs for natural language tasks and AMX verifies every output — no unverified decisions reach your systems.
• Process modeling — invoice-to-payment, regulatory reporting, supply chain management, order fulfillment — modeled and verified end-to-end
• Agent protocols — Coworker agents communicate via MCP (Model Context Protocol), A2A (Agent-to-Agent, Linux Foundation standard), and WebMCP (W3C path) for interoperable, standards-based orchestration
• Verified execution — when Coworker agents act on real systems, every decision is backed by an XFPC proof certificate. No unverified AI decisions in your operations.
• Native apps — Coworker agents run natively on iOS, Android, macOS, Windows, and Linux

Verification Engine

The core of the AMX platform. A symbolic reasoning engine that re-derives AI answers from source evidence through deterministic logic.

• Independent re-derivation — AI answers are verified through a completely separate computational path, not by checking the AI’s work, but by solving the problem independently from source evidence
• Zero neural components — the verification path contains no neural networks, no language models, no probabilistic inference. It cannot hallucinate.
• Deterministic — same input produces the same output and the same proof, every time. No randomness, no variation between runs.
• Fast-path verification — 10ms for known patterns and cached derivations. 500ms for full derivation of complex queries requiring complete reasoning chains.
• Pre-production error catching — discrepancies between AI output and verified output are flagged before they reach production systems

XFPC Proof Certificates

Every verified decision produces a proof certificate following the eXtensible Formal Proof Certificate (XFPC) open standard.

• Open standard — XFPC is not a proprietary format. Any system that implements the standard can verify proofs independently.
• Independently replayable — any auditor can verify the proof on their own hardware, at any time, without access to AMX systems
• Tamper-evident — cryptographic integrity ensures proof certificates cannot be altered after issuance
• Legally admissible — designed for legal evidence under the Daubert standard. Satisfies explainability requirements of the EU AI Act (Article 13), MAS guidelines, and SEC/OCC expectations.
• Full provenance tracking — every proof links to the exact Knowledge Capsules and evidence that contributed to each decision
• Archivable — proof certificates are self-contained records suitable for long-term regulatory retention and Freedom of Information requests

Knowledge Capsule Workbench

The complete toolchain for authoring, testing, linking, and publishing Knowledge Capsules — the modular knowledge packages that power AMX verification.

Authoring

• Kotlin DSL — Knowledge Capsules are authored using a purpose-built Kotlin-based domain-specific language. Domain experts and engineers define verification logic, business rules, and regulatory constraints in a structured, type-safe format.
• Cryptographic signing — every published capsule is cryptographically signed. Verification consumers can confirm capsule authenticity and integrity before use.
• Training and inference tools — build capsules from organizational documents, regulatory texts, and domain knowledge. The workbench supports both supervised capsule construction and automated knowledge extraction.
• Interactive testing — test capsules against sample inputs, inspect derivation paths, and validate correctness before publishing via amx cps test and amx cps validate.

Linking and Composition

• Capsule linking — Knowledge Capsules can be linked and associated with other capsules. A trade compliance capsule can reference a tariff classification capsule, which in turn links to a currency conversion capsule. The system guarantees consistency across linked capsules.
• Dependency management — capsules declare their dependencies explicitly. The workbench validates that all dependencies are satisfied and consistent before publishing.
• Versioned composition — combine capsules from different domains and versions. The platform tracks which capsule versions contributed to each verification.

Publishing

• Maven repository publishing — Knowledge Capsules are publishable to any Maven-compatible repository. Use your existing artifact infrastructure (Nexus, Artifactory, Maven Central, or private repositories) to distribute and version capsules.
• Semantic versioning — capsules follow semantic versioning conventions. Breaking changes, additions, and patches are tracked automatically.
• Hot-swappable deployment — published capsules can be deployed to running AMX instances without downtime. Update a capsule and the verification engine picks up the new version immediately. No retraining, no redeployment cycles.

Domain-Specific Capsules

• Pre-built capsules for finance, legal, healthcare, manufacturing, and trade compliance
• Malaysian regulatory capsules: LHDN, SSM, MyTrade, PDPA
• Industry certification capsules: HALAL, ISO, HACCP

Multilingual Intelligence

Multilingual support designed for global and Southeast Asian business environments, with tiered production readiness.

• Production-grade — English has full verification support across all Knowledge Capsules and document types
• Strong support — Chinese (Mandarin) and Malay (Bahasa Malaysia) have robust support for business document processing, financial workflows, and compliance verification
• Foundational support — Tamil has foundational coverage with ongoing expansion
• Cross-language verification — understand in one language, verify and answer in another, with proof integrity preserved across languages
• Extensible architecture — designed for rapid addition of new languages without rebuilding the verification engine

Deployment Options

Integration

AMX is designed to fit into your existing infrastructure with minimal friction.

• Sidecar mode — deploy alongside your existing AI systems. Zero code changes to your LLM. AMX verifies outputs transparently.
• API gateway mode — centralized verification governance. All AI outputs pass through AMX before reaching downstream systems.
• Embedded mode — for low-latency requirements. 10ms fast-path verification for known patterns and cached derivations.
• Asynchronous mode — post-hoc audit and compliance verification. Verify historical AI decisions for regulatory review.
• Agent protocol support — MCP, A2A, and WebMCP for standards-based integration with agent ecosystems
• Enterprise connectors — integrates with existing ERP, CRM, and business systems. Works within your current technology stack.

Platform Economics

Enterprise AI spending has crossed $200 billion annually, yet only 30-40% of AI project value is realized — the gap driven by trust deficits, manual review overhead, and regulatory constraints. AMX closes this gap with economics that make verification a no-brainer:

Security and Compliance

• Zero data egress — verification happens where your data lives. Nothing leaves your infrastructure unless you choose otherwise.
• SOC 2 and ISO 27001 — aligned security practices across the platform
• Data protection compliance — PDPA (Malaysia), GDPR (EU), PIPL (China) compliant deployment configurations
• EU AI Act readiness — XFPC proof certificates satisfy explainability and auditability requirements for high-risk AI applications
• Role-based access control — granular permissions for verification policies, Knowledge Capsules, and proof certificate access
• Complete audit logging — every verification, every capsule update, every access event is logged and immutable

Coming Soon

AMX Central — a marketplace for verified Knowledge Capsules. Discover, share, and deploy domain-specific capsules across your organization and with partners. Think “GitHub for enterprise knowledge.” Capsule authors keep 30% of revenue. Quality signals, usage metrics, and verification status help consumers find the right capsules.

Knowledge Refinery — automatically transform organizational documents, policies, and procedures into verified Knowledge Capsules. A hybrid pipeline where LLMs extract knowledge and AMX verifies correctness — solving the cold-start problem for organizations with existing institutional knowledge locked in documents, manuals, and policies.

Knowledge Copilot — an interactive assistant that helps domain experts author, test, and refine Knowledge Capsules without requiring engineering skills. Guided capsule creation with the Kotlin DSL — domain experts focus on what they know, the Copilot handles the technical structure.

Coworker Marketplace — discover and deploy pre-built verified Coworker agent workflows for common business processes. Pre-verified, ready to deploy, backed by XFPC proof certificates from day one.

The “Intel Inside” for AI

AMX does not compete with LLMs — we make every LLM trustworthy. Just as Intel provided the trusted processing layer inside every computer regardless of manufacturer, AMX provides the trusted verification layer inside every AI system regardless of model. Our platform works with GPT-4, Claude, Llama, Mistral, Gemini, or your own fine-tuned models. The AI landscape will continue to evolve rapidly; the need for verification is permanent.