Custom AI Agent Engineering

AI engineering,
on solid foundations.

We build custom AI agents, MCP servers, and RAG pipelines on top of the production software systems we've been shipping for a decade — without breaking the architecture you depend on.

10+
Years shipping production software
24/7
Government & platform-grade systems
AI-native
Small team, fully automated delivery
PLAN
Decompose goal
TOOL
Call MCP / API tool
OBSERVE
Parse structured output
REFLECT
Update state, re-plan
AGENT LOOP
Skills · Tools
Memory · Plan
agent.runtime/v2
live
OUR APPROACH

Most teams know they need AI.
Few know exactly what to build.

Writing a perfect spec from scratch is hard. Most projects stall there.

We flip it: a working prototype within days, in front of you. You react, we listen, we adjust — and the real requirements surface naturally from the thing you can actually click on. By week three, what you actually need is on the screen.

  • Discovery call → working prototype in 5–10 days
  • Iterate on something real, not a 40-page spec
  • Final scope emerges from feedback, not guesswork
  • You own the code from day one
DISCOVERY → DELIVERY~3 WEEKS
  1. 1
    ListenDay 1

    30-min call. We map your domain, not a template.

  2. 2
    PrototypeDay 5

    A clickable agent runs against your real data.

  3. 3
    ReactDay 10

    You use it. We watch where you stumble.

  4. RefineDay 14

    Real requirements emerge from real friction.

  5. ShipDay 21

    Production-grade integration, not a demo.

ABOUT

Production software engineers,
now building for the AI era.

X CODE ONE is a Kuala Lumpur–based engineering studio. For years we've shipped government-grade platforms, IoT systems, and the kind of distributed services that can't afford to fall over.

Today, we're a small team running fully AI-automated delivery. The same people who care about thread safety and migration paths now build the agent runtimes, MCP servers, and RAG pipelines that sit on top of your existing software — without compromising it.

That's the bet: you get AI velocity without throwing away the architectural discipline that keeps your business running.

Headquarters
Kuala Lumpur, Malaysia
Founded
2024 (re-org from 10+ yrs of engineering)
Team
Small, fully AI-automated
Clients
Government · Platforms · In-house products

Don't break what works

AI bolted onto fragile foundations is a future incident report. We integrate, we don't replace.

Boring architecture, sharp AI

Vibecoding scales until it doesn't. We use proven patterns under the agents.

Show, don't spec

Working prototypes uncover real requirements faster than any document.

Yours, top to bottom

Your code, your auth, your data. No proprietary runtime lock-in.

SERVICES

From custom agents to system-wide AI integration.

01

Custom AI Agents

Domain-specific agents with curated skills, tools, memory, and reflection loops — designed around your business, not a generic chatbot.

  • Skill / tool / agent-loop design
  • Stateful memory + recovery
  • Observability & guardrails
02

MCP Servers · API → Tools

We wrap your existing APIs into typed, audited Model Context Protocol tools an LLM can call safely under your auth.

  • MCP server implementation
  • Schema + permission gating
  • Backwards-compatible adapters
03

RAG · LangChain · LLM Analytics

Retrieval pipelines over your databases and documents — chat with millions of rows, generate statistics, surface anomalies.

  • Vector + hybrid retrieval
  • Streaming SQL-on-LLM
  • Evaluation harness
04

Safe Integration with Legacy

Bring AI into systems that absolutely cannot break. We preserve your domain layer, transactions, and audit trail.

  • Non-invasive adapters
  • Sidecar deployment patterns
  • Rollback-first delivery
05

Web · Mobile · IoT

The original X CODE ONE — full-stack web platforms, hybrid mobile, and IoT control + telemetry, delivered to government-grade SLAs.

  • .NET / Node / Python
  • Angular / Ionic / React
  • Real-time IoT + dashboards
06

Consulting & Architecture

Not sure where to start? A short paid engagement to map your domain, score candidate AI surfaces, and price the prototype.

  • AI readiness assessment
  • Architecture review
  • Build-vs-buy guidance
ARCHITECTURE

AI on top. Your foundations untouched.

We treat your existing services, databases, and message buses as the source of truth. AI sits as a new tier — typed, audited, and removable — calling down through the same domain layer your team already trusts.

AI TIER · NEW
Agent Runtime
Skills · Tools · Loop · Memory
MCP Servers
API → typed tools
RAG / LangChain
Vector + Hybrid
LLM Gateway
Routing · Eval · Safety
DOMAIN · PRESERVED
Domain Services
.NET · Python · Node
AuthN / AuthZ
OIDC · RBAC · Audit
Workflow / Jobs
Sagas · Schedulers
INFRA · PRESERVED
RabbitMQ
Async messaging
Redis
Cache · Pub/Sub · Inter-svc
Socket.io · TCP/IP
Realtime sockets
Docker · Azure
Containers · Cloud
DATA · PRESERVED
MSSQL
OLTP · Reporting
PostgreSQL
OLTP · JSONB
MySQL
Legacy systems
Vector Store
pgvector · Qdrant
AGENT ANATOMY

What goes inside a serious agent.

SKILLS
Domain skills
Procedural know-how the agent can rely on — written by us, reviewed like any code.
TOOLS
Typed tools
Each tool is a typed function backed by your real services and databases.
LOOP
Reasoning loop
Plan → call tool → observe → reflect. Bounded, observable, recoverable.
MEMORY
State & memory
Short-term scratchpad, long-term store, vector recall — all auditable.
TECH STACK

Database-agnostic, language-fluent, architecture-disciplined.

We don't ideologize the stack. MSSQL, MySQL, Postgres — all production-ready for us. Python and Node sit comfortably alongside .NET. Redis and RabbitMQ are how our services talk; Socket.io and raw TCP/IP sockets are how they go realtime.

Frontend
AngularReactIonicHTML5CSS3TypeScript
Backend
.NET CoreC#PythonNode.jsEntity Framework
AI
LangChainMCPRAGVector DBsLLM GatewaysAgent Runtimes
Realtime & Messaging
RabbitMQRedisSocket.ioTCP/IP raw socketWebSocket
Databases
MSSQLPostgreSQLMySQLpgvectorQdrant
Infra & Cloud
DockerAzureCI/CDObservabilityIoT Edge
CONTACT

Tell us a rough idea. We'll show you a prototype.

Free 30-minute consultation. No NDA needed for the first call. We'll come back within 48 hours with a one-page direction and a prototype timeline.

Emailhello@xcodeone.com
OfficeKuala Lumpur, Malaysia
HoursMon–Fri · 09:00–18:00 MYT
ResponseWithin 48 hours, on business days
Not sure what to build yet? That's the most common situation. Book the call anyway — we'll help you find the highest-leverage AI surface in your existing systems.
We respond within 48h · No spam · No mailing list
X CODE ONE · Coding True Light into Life
© 2026 X CODE ONE SDN. BHD. · Kuala Lumpur, MY · v0.0.1