HomePortfolioTeamBlogSearch
AI & Agentic AI
AI Consulting and Implementation Gen AI Chatbot Agentic AI with N8N Anthropic Claude Partner Malaysia: AI Development for Enterprise By Claude Certified Architect (CCA-F) Claude AI Training Malaysia (Technical): CCAR-F Preparation Class Claude AI Training Malaysia (Non-Technical): CCAO-F Preparation Class Claude Customer Story: Hernan Corporation puts a Claude agent in every stage of its durian cold chain Claude Customer Story: Zetrix ships the agent interface to its Layer-1 blockchain with Claude Code OpenAI ChatGPT Partner Malaysia: Enterprise AI Solutions Openclaw for Enterprise SMEHero.ai: AI Platform for Malaysia SMEs Gen AI OPC One Person Company AI Forward-Deployed Engineers FDE as a Service Enterprise Voice AI Agent GPU-as-a-Service (GPUaaS) in Malaysia Malaysia GPU as a Service Guides: From GPU, DC, Power, Software, KYC to Pricing AI Gateway Malaysia | LiteLLM, Bifrost, Quota Management and Enterprise AI Governance AI and Data for General Elections ReelHero AI: AI Templated Video Generator Gen AI Digital Avatar AI Digital Banking Malaysia Conversational Payment Solutions AI Chatbot Automated Testing: LLM as a Judge and Red-Teaming Malaysia AI Vibe Coding/VibeOps LLM Fine-Tuning as a Service MySTI-Certified AI Tech Products AI Deepfake Detection
Development & Apps
Our Services AR/VR/XR Apple Vision Pro Development Blockchain/Web3/Smart Contract UI/UX Design Backend Development Power BI Dashboard EV Solutions – OCPP/OCPI/E-MSP Web App Strapi Headless CMS Development SuperApp Development Flutter React Native Native iOS Native Android CarPlay/Android Auto In Car Assistant
Cloud, Fintech & Enterprise
AWS Malaysia Cloud Modernization: Your Trusted AWS Partner Use Case: JomeInvoice Use Case: DahReply Azure Malaysia Cloud Modernization: Your Trusted Microsoft Partner Malaysia Cloud Modernization with Local BytePlus Cloud: Your Trusted BytePlus Partner Lark Malaysia Reseller For SME and Corporates Whitelabel E-Wallet Development Regulated Fintech Development Clean Energy Digital Solutions Malaysia Malaysia IOT Platform/Hardware Integration Software Development MyDigital ID Integration Property Development Digital Platforms Malaysia Cloud Framework Agreement CFA for Government Enterprise Grade IRB/LHDN E-Invoicing Middleware Meta WhatsApp and SMS API Services
Team Extension / Staff Augmentation
Team Extension Java Springboot Developer Staff Augmentation and Outsourcing Rust Developer Staff Augmentation and Outsourcing Golang Developer Staff Augmentation and Outsourcing C# .NET Core Developer Staff Augmentation and Outsourcing MuleSoft Developer Staff Augmentation and Outsourcing Gen AI LLM Developer Staff Augmentation
More
Press Our Standards – CMMI and ISOs Certified Join Agmoians Project Management Enterprise Grade Test Automation For Software Applications Coding Training Blog Government Grant Outsystems Agmo Singapore Agmo Academy
Request a Quote/Demo
CLAUDE CUSTOMER STORY

Claude Customer Story: Zetrix ships the agent interface to its Layer-1 blockchain with Claude Code

Zetrix runs its national blockchain infrastructure on an agent interface built with Claude Code

55 toolsin the production Zetrix MCP Server — published to npm, running against mainnet
91 of 91commits in Zetrix’s two shipped MCP servers co-authored by Claude
1 dayfrom empty repository to the first production release on npm
6 componentsin production across Zetrix’s live blockchain estate

Most blockchains are judged by what happens on the chain. Zetrix is judged by what happens at the border.

Operated by Zetrix AI Berhad — the Malaysian company formerly known as MY E.G. Services, the country’s long-running e-government concessionaire — Zetrix is a Layer-1 public blockchain that has been live on mainnet since April 2022, running across more than 300 nodes in 20 countries. It is not a test network or a research chain. It operates the Xinghuo BIF International Supernode connecting China’s national blockchain infrastructure to markets outside it. It is the protocol under Malaysia’s national blockchain infrastructure, built with the national R&D agency MIMOS and activated in April 2025, and under the Philippines’ national public blockchain. It carries live customs and trade documentation between Malaysia, China and the Philippines for the General Administration of Customs of China, Malaysia’s Ministry of Investment, Trade and Industry, and the Philippine Bureau of Customs. And it issues W3C-standard decentralised identifiers and verifiable credentials that let one country’s agency check another country’s document without either side handing over the underlying data.

That production footprint created a distinctive engineering problem, and in 2025 it got sharper. The parties that most need to touch this infrastructure are no longer only developers. They are AI agents — asked to verify a bill of lading, issue a credential, or pay for a service and prove who authorised the payment. Zetrix needed a supported, production-grade way for them to do it.

The challenge

Making a live chain agent-operable is not a thin wrapper. It is a large surface of exacting, specification-bound code that has to be correct against real value: the full RPC and WebSocket surface, protobuf message handling, transaction construction with dynamic fee estimation, eight distinct cryptography operations, Ed25519 signing and BBS+ selective disclosure, HMAC-authenticated gateway clients, and — the part with no analogue in ordinary SDK work — tool descriptions written carefully enough that an agent reaches for the right operation and never asks a human for a private key.

Zetrix needed that surface in several places at once, all of it destined for production rather than a lab. A general server for the chain itself. A second for the verifiable-credential service, because digital identity is the product governments actually buy. A third for TradeTrust electronic bills of lading on Zetrix’s L2, because cross-border trade is where the chain earns its keep. And underneath all of it, payment rails for agents: the x402 protocol for machine payments over HTTP, and x401, Zetrix’s own complement to it — where x402 asks “can you pay?”, x401 asks “who authorised this, and are they allowed?”

The approach

Zetrix put Claude Code at the centre of the work and adopted it across the engineering organisation on a Claude team plan. The choice is legible in the company’s own public record: rather than treating AI assistance as an invisible convenience, Zetrix’s engineers left the attribution in place. Every commit in the two flagship MCP servers carries a Co-Authored-By trailer naming the model that helped write it — Claude Opus 4.6 running in its 1M-context configuration on the large majority, with Claude Sonnet 4.6 and Claude Sonnet 5 on later work.

The pattern across those repositories is compressed first releases followed by fast, surgical maintenance against live traffic. The Zetrix MCP Server’s entire first production release — 55 tools across seven categories, dual stdio and HTTP transports, a full tool catalogue, developer guides and a test report — landed in seventeen commits on a single day in November 2025 and was published to npm the same day. The Zetrix VC MCP Server went from empty repository to twenty tools spanning DID generation and resolution, credential issuance, revocation and selective-disclosure presentations in three days the following April, with thirty-two of its thirty-seven commits on day one.

The maintenance commits are as telling as the releases, because they are the record of software being run against live traffic. “Strengthen tool descriptions to not ask user for credentials.” “Strengthen VP tool routing so agents don’t pick vp_create for ‘create VP’.” “Fix decrypt failing when encryptedData passed as JSON string.” “Fix SDK using HTTP instead of HTTPS for mainnet.” This is the texture of operating agent-facing infrastructure: the hard part is not the happy path but the hundred small ways a tool surface misleads the model calling it — and it is work that goes faster when the thing helping you write it is the same kind of thing that will consume it.

In production

Every component described here is deployed, released and in daily operational use.

The results

What it unlocked

The strategic point is not that Zetrix wrote software faster. It is what the software is for. Zetrix’s position is that the next decade of cross-border trade and digital identity will be transacted by agents acting on behalf of institutions, and that those agents will need to prove authorisation before they are allowed to act — the problem x401 exists to solve, on top of an identity stack Zetrix already runs for national governments. Building that layer with Claude Code let the company move at the speed the market is moving rather than the speed of a hiring plan.

In February 2026 the International Finance Corporation took a stake in the company to fund blockchain-based digital public infrastructure and specialised agentic AI models. In April, Zetrix and CAICT’s Astron launched a blockchain trust layer for autonomous AI agents. In July, an MOU made Zetrix the protocol under the Philippines’ national public blockchain, with the Department of Information and Communications Technology, the Bureau of Internal Revenue, the Securities and Exchange Commission and the Philippine Ports Authority named among the agencies it will serve. The MCP servers are the connective tissue between all of it and the agents that will use it.

“A blockchain is only as useful as the software people can reach it with. Claude Code let a small team write the whole agent-facing surface of Zetrix — tools, SDKs, docs, tests — and ship it as production infrastructure that governments depend on.”

— Rushdan Anuar, Chief Technology Officer, Zetrix