Skip to content
Enterprise AI · Multi-cloud · Singapore

Build agents.Ship to production.

Two products, one platform. xmCoder builds enterprise AI agents in a sandbox. xmGate routes them across AWS Bedrock, Azure OpenAI, and Google Vertex.

19Models
2Clouds
SOC 2Compliant
99.9%Uptime SLA
One Platform · End to End

From your trigger to the right cloud — in one hop.

xmCoder runs your sandboxed agent. xmGate routes every LLM call across AWS, Azure and Google. One stack, one contract, one bill.

Agent Builder · Enterprise
xC

xmCoder

Build, sandbox-test and deploy AI agents in one tool. Visual canvas for non-engineers, full code escape hatch for everyone else. Publish as a web app, REST API or embedded widget — with a single command.

  • Sandboxed agent execution — strict resource & network limits
  • Visual canvas + code — TypeScript, Python or natural language
  • One-click publish — web app, REST API or embedded widget
  • Enterprise SSO (SAML/OIDC) + tamper-evident audit log
  • Custom guardrails, memory and tool registries
~/projects/api-server
Gateway · Multi-Cloud LLM Routing
xG

xmGate

A curated gateway to every major cloud LLM. xmGate brings AWS Bedrock, Azure OpenAI, Google Vertex AI and direct provider APIs behind one battle-tested endpoint — drop-in compatible with your existing OpenAI / Anthropic / Google SDK calls.

  • 21 frontier LLMs across AWS Bedrock, Azure OpenAI, Google Vertex + direct APIs
  • VPC peering & PrivateLink on AWS, Private Endpoints on Azure
  • IAM / managed-identity / service-account auth — no static keys
  • Unified observability: CloudWatch, Azure Monitor, Cloud Logging
  • Drop-in compatible with OpenAI / Anthropic / Google SDKs
POSTapi.xmai.sg/v1/chat/completions200
route: aws/bedrock · azure · vertex42ms
picked → bedrock/claude-opus-4.7OK
AWS47%
Azure31%
GCP22%
Enterprise-Grade Foundations

Built for production at scale from day one

Sandboxed execution, multi-cloud plumbing, enterprise auth — the boring infrastructure work, already done.

Sandboxed execution

Every agent runs in an isolated runtime with strict CPU, memory and network limits. No access to your prod data unless you grant it explicitly via signed connectors.

Build your own agents

Visual canvas for non-engineers, TypeScript / Python escape hatch for power users. Compose tools, prompts, memory, guardrails and human-in-the-loop steps.

One-click publish

Ship as a hosted web app, public REST endpoint, or embeddable widget. Custom domain, SSL and auth wall provisioned automatically — zero infra to manage.

Multi-cloud routing

Route across AWS Bedrock, Azure OpenAI, Google Vertex AI and direct provider APIs (OpenAI, Anthropic, Mistral, Cohere). Pick any cloud per request — one auth, one billing, one observability stack.

Enterprise SSO + audit

SAML / OIDC, role-based access, project-scoped permissions. Tamper-evident audit log of every agent action, prompt, response and tool call. SOC 2 Type II in progress.

Auto-failover

Provider down? Request is rerouted to the next healthy endpoint in the same model family with zero application changes. Failover SLO surfaced in CloudWatch.

Supported Models

Every frontier LLM, one integration.

21 models across 3 providers, refreshed to April 2026. Swap models freely without changing a line of client code.

OpenAI

9 models
gpt-5.5gpt-5.5-progpt-5.4gpt-5.4-minigpt-5.4-nanogpt-5.4-progpt-5.2gpt-5.2-progpt-5
Enterprise Security & Compliance

Built for the CISO review

Per-cloud tenant isolation, cloud-native auth (no static keys), tamper-evident audit logs and zero data retention. SOC 2 Type II in progress, GDPR / PDPA / ISO 27001 already covered.

SOC 2 Type II in progress

Independent audit underway with completion targeted Q3 2026. ISO 27001, GDPR and PDPA compliance already in place. Available under NDA on request.

Tenant isolation per cloud account

Every enterprise tenant gets a dedicated AWS account, Azure subscription or GCP project boundary. No shared compute, no shared database. Cross-tenant data access is structurally impossible.

Zero data retention

Prompts and model responses pass through and are never persisted. Once the request completes, the content is gone — auditable in CloudWatch flow logs.

Never used for training

Your API traffic is never used to train, fine-tune or improve any AI model — ours or anyone else's. Your IP and customer data stay 100% yours.

Cloud-native auth, no static keys

TLS 1.3 in flight, AES-256 at rest. Authenticate via AWS IAM roles, Azure Managed Identities or GCP Service Accounts — short-lived tokens only, never long-lived API keys.

Tamper-evident audit log

Every agent action, prompt, response and tool call is recorded with a hash chain. Stream to your SIEM (Splunk, Datadog, CloudTrail) via EventBridge.

For procurement & security teamsFull audit pack — SOC 2 progress letter, pen-test summary, DPA templatesecurity@xmai.sg
Frequently Asked

Procurement, security & platform questions

The answers we keep giving on enterprise sales calls — written down.

Bedrock gives you AWS-only model access. xmGate gives you one API across AWS Bedrock, Azure OpenAI, Google Vertex and direct providers — same auth, same billing, same observability — plus auto-failover when a provider hiccups. xmCoder layers a sandbox + publish workflow on top, so engineering teams aren't gluing five SDKs together.

Ready to ship?

Start with xMAI today

Build, test and deploy AI agents with xmCoder. Route every LLM call across AWS, Azure and Google with xmGate. Enterprise-ready from day one — one platform, one contract, one integration.