Create reliable AI agents
for daily challenges

Centrix grounds large language models in business logic for chat, desktop operations, and visual public-cloud agents.

Create Agent Select a capability, then deploy.

Metro-Suisse Phase II is on schedule for Q4 2026. Compute fleet scaled to 12 instances.

What is the current cloud spend vs budget?

Ask the agent…

An AI solution tailored to your business needs.

The Centrix platform
Capabilities

Three agents. One operating layer.

Centrix ships multi-function agent technology — not a chatbot wrapper. Each agent sees, decides, and acts inside a governed runtime.

Ticket #4821 classified as billing. Drafting a reply with policy 4.2.
Approve and send.

01 — Conversation

Chat Agent

Grounded dialogue for support, ops, and internal knowledge. Business rules, tools, and audit trails sit beside the model — so answers stay on-policy.

  • Policy-aware replies and tool calling
  • Human-in-the-loop handoff
  • Session memory with redaction
Operations Desk
invoice_sep.xlsxopen
vendor-formfill
SAP GUIwait

02 — Desktop operations

Desktop Agent

A visual operator for the desktop. It reads the screen, moves the cursor, types into native apps, and completes back-office work that APIs never reached.

  • Screen understanding + click / type / drag
  • Works across legacy and modern apps
  • Step-level logs for every action
Instances · ap-southeast-1
cx-api-01runningc6i.large
cx-etl-02runningr6i.xlarge
cx-gpu-1stoppedg5.xlarge
Agent: applying security group sg-edge

03 — Public cloud

Cloud Visual Agent

A visual proxy that operates public-cloud consoles the way an engineer does — seeing the UI, clicking through, and executing change with a recorded trail.

  • Console-level visual control (AWS / Azure / GCP)
  • Provision, scale, secure, and audit
  • Guardrails before destructive actions
The platform

Business logic around the model — not inside a prompt.

Centrix wraps every agent with policy, tools, memory, and evidence. The model reasons. The platform decides what is allowed to happen.

  1. Perceive Chat tokens, desktop pixels, or a live cloud console — unified as observations.
  2. Ground Retrieve policy, CMDB, runbooks, and prior actions before the next move.
  3. Act Reply, click, type, or call an API — with confirmation on high-risk paths.
  4. Prove Every session leaves a replayable trail for security, finance, and ops.
Observation · Policy · Action · Evidence
Use cases

Where agents earn their keep.

Customer operations

Chat agents that know your SKUs, SLAs, and refund policy — and escalate with full context.

Back-office desktop

Invoice capture, ERP posting, and portal updates on software that has no API.

Cloud operations

Visual agents that execute runbooks in the console: scale events, IAM hygiene, spend checks.

Hybrid workflows

A chat request that becomes a desktop fill, then a cloud change — one session, three surfaces.

Live preview

Try an agent path.

A guided simulation of the three surfaces. No account required.

Company

CENTRIX TECH PTE. LTD.

A Singapore technology company building multi-function AI agent infrastructure for enterprises that need chat, desktop, and public-cloud visual operations in one governed stack.

Book a demo

See the agents on your workflows.

Tell us whether you need chat, desktop operations, cloud visual control — or all three. We will schedule a working session from Singapore.

This preview stores the request on this device. Production wiring can post to your CRM.