Hosted in Sydney, Australia

Private AI agents,
run locally in Sydney

Deploy a fully managed OpenClaw instance with local model inference, KVM-level isolation, and Australian data sovereignty. We run the model stack ourselves on Sydney GPU capacity, keep prompts off third-party model APIs by default, and give every tenant their own secure runtime.

clawd.au

$ openclaw status

Agent: ClawHead

Status: Running

Region: AU-SYD

Inference: Local gpt-oss-120b

Facility: Equinix Sydney

Backups: Daily, encrypted

$ _

Sydney-hosted

Models, tenants, and platform services run locally in Sydney.

Tier IV footprint

Hosted in Equinix Sydney facilities with enterprise-grade controls.

Prompt privacy

We do not centralise prompt logs. We keep metrics and usage instead.

Encrypted backups

Daily backups are stored off site in encrypted form.

Features

Built for private workloads,
not token resale

Same secure operating model on every plan.

01

Unlimited local inference

We run gpt-oss-120b ourselves on Sydney GPU capacity, so you are not paying per prompt or watching token counters all day.

02

Hard tenant isolation

Every tenant runs in its own microVM via Kata Containers and Cloud Hypervisor. Separate kernel, separate memory, separate runtime boundary.

03

Australian data residency

Application services, tenant workloads, model serving, and backups stay in Australia. Your default path never leaves the country.

04

Secure access path

Traffic enters through a dedicated auth gateway with TLS at the edge and trusted-proxy headers enforced before requests reach your tenant.

05

No prompt logging

We track operational metrics and usage for billing and reliability. We do not centralise prompt logs across the platform.

06

Optional BYO provider keys

If you decide to connect an external model provider, you can bring your own API key. That choice is optional and stays under your control.

How We Run AI

We rent GPUs locally,
then run the stack ourselves

Clawd.au is not a thin wrapper over overseas inference APIs. We rent GPU capacity from Sydney providers, run the model stack locally, and operate the serving path ourselves. That is what makes unlimited inference possible without turning your prompts into someone else's SaaS telemetry.

If you want to use an external model provider, you can bring your own API key. That traffic is your choice and your responsibility. The default Clawd.au path stays on Clawd-managed Australian infrastructure.

Local GPU capacity

We source compute in Sydney so model execution happens close to the rest of the platform.

Platform-controlled runtime

We own the inference path, networking, and model serving stack instead of reselling token APIs.

Predictable pricing

Unlimited inference works because pricing is built around infrastructure capacity, not per-request markups.

Optional external providers

BYO API keys are supported when you want them, but they are not the default operating model.

Security

Secure by default,
isolated by design

We treat each tenant as a separate environment. Isolation, gateway enforcement, encrypted transport, and backup discipline are part of the platform architecture, not optional add-ons.

  • Dedicated microVM per tenant with KVM-level isolation
  • Network isolation between tenants and controlled ingress paths
  • Auth gateway plus trusted-proxy mode before requests reach OpenClaw
  • TLS-secured access and encrypted off-site daily backups

Prompt privacy

Prompt text is not copied into a central logging pipeline. We record usage and platform metrics instead.

Tenant boundaries

Each tenant runs in its own microVM with isolated compute, storage, and network policy boundaries.

Secure gateway

Requests are authenticated at the edge, then forwarded through a controlled gateway with trusted headers.

Backup discipline

Daily backups are stored off site in encrypted form so recovery does not depend on a single facility.

Data Sovereignty

Sovereign infrastructure,
all the way down

Clawd.au runs on Australian infrastructure end to end. Tenant workloads, model execution, storage, and platform services stay local, with production hosting in Equinix Sydney facilities.

  • Models and tenant workloads run in Sydney, not offshore
  • KVM-isolated microVMs per tenant
  • Daily off-site backups stored in encrypted form
  • No third-party AI API calls by default — prompts stay on Clawd.au infrastructure

AU

Sovereign

Pricing

Pick the size,
not the privacy model

Every plan gets the same local inference model, the same security posture, and the same sovereign hosting. The difference is capacity.

Lite

$19 /mo

  • 2 CPU cores, 2 GB RAM, 20 GB disk
  • Unlimited local gpt-oss-120b inference
  • WebChat and channel support
  • KVM-level isolation
  • Daily encrypted backups
Get started

Pro

Popular

$39 /mo

  • 4 CPU cores, 4 GB RAM, 40 GB disk
  • Unlimited local gpt-oss-120b inference
  • WebChat and channel support
  • KVM-level isolation
  • Daily encrypted backups
Get started

Max

$79 /mo

  • 8 CPU cores, 16 GB RAM, 100 GB disk
  • Unlimited local gpt-oss-120b inference
  • WebChat and channel support
  • KVM-level isolation
  • Daily encrypted backups
Get started
See all plan details

Need to call an external model provider? Bring your own API key. That traffic is optional and remains under your control.

FAQ

Questions about trust, privacy, and scale

How can you offer unlimited inference?
We price around infrastructure capacity, not token resale. Clawd.au runs the default model stack locally on Sydney GPU capacity we rent and operate ourselves.
Do you send prompts to model providers?
No, not by default. Prompts stay on Clawd.au-managed Australian infrastructure unless you explicitly configure your own external provider with your own API key.
Do you log prompts?
We do not centralise prompt logs for platform analytics. We keep service metrics and usage data needed for billing, reliability, and operations.
Where is the platform hosted?
Clawd.au is hosted in Sydney, including the model runtime and tenant workloads, with production infrastructure in Equinix Sydney facilities.
How are tenants separated from each other?
Each tenant runs in its own microVM with KVM-level isolation, separate storage, and network controls so one tenant does not share a runtime with another.
How do backups work?
We take daily backups and store them off site in encrypted form so recovery is possible without exposing tenant data in the process.

Ready to deploy your
AI agent?

Sign up in under a minute. Your OpenClaw instance will be live on secure Sydney infrastructure before your coffee gets cold.

Get started — from $19/mo