Shared
Best for teams getting started with Operario.
Pricing to be announced
- Operario in our cloud, operated by us
- Nothing to install or maintain
- Automatic updates
- Standard support
- 90 days of task logs by default
Which one is yours?
Start from your situation, not from a feature list.
Trying Operario? The fastest way to put it to work.
Shared →Need your own deployment, without running infrastructure?
Dedicated instance →Answering to auditors or regulators?
Dedicated account →Nothing may leave your premises?
Self-hosted →Dedicated instance
Best for teams that need dedicated performance and their own deployment.
Pricing to be announced
- Your own deployment, operated by us
- Dedicated services and an isolated task store
- Dedicated subdomain
- Updates and maintenance windows you agree
- Priority support
Dedicated account
Best for regulated teams and enterprises that need account-level separation.
Pricing to be announced
- Your own AWS account in our cloud, operated by us
- Your own identity, network, keys, and audit trail
- Dedicated subdomain
- Updates and maintenance windows you agree
- Priority support
Self-hosted
Best for organizations whose systems must stay on their own infrastructure.
Pricing to be announced
- You run Operario in your own cloud or on your own hardware
- Your domain, your data, your keys
- New images; you apply updates
- Support contract, optional managed setup
- Retention entirely yours to set
45 of 56 tasks · 3rd of 25 teams — Frontier AI Builders Hackathon. Screenshots are never stored by us, in any tier: they travel to the vision model in memory and in transit only.
Compare tiers
The differences that decide it. Everything else is the same Operario.
Scroll sideways to compare all tiers.
| Shared | Dedicated instance | Dedicated account | Self-hosted | |
|---|---|---|---|---|
| Operated by | Us | Us | Us | You |
| Runs in | Our cloud | Own deployment | Own AWS account | Your hardware |
| Isolation | Tenant-isolated | Dedicated deployment | Dedicated account | Your infrastructure |
| Address | app.operario.tech | Dedicated subdomain | Dedicated subdomain | Your domain |
| Updates | Automatic | Scheduled, agreed | Scheduled, agreed | You apply |
| Maintenance windows | Shared rollout | Agreed | Agreed | Yours |
| Support | Standard | Priority | Priority | Support contract |
| Extended retention | Paid add-on | Paid add-on | Paid add-on | Yours to set |
Add-ons
Available on top of any tier that fits.
Extended retention
Task logs are kept 90 days by default in every operated tier. Longer horizons — 1, 3, 5, or 7 years — hold the full task record to the chosen horizon.
A paid add-on in Shared, Dedicated instance, and Dedicated account. Self-hosted customers control retention themselves.
Bedrock regulated inference
For regulated tenants, the vision model runs through AWS Bedrock instead of a third-party AI provider: inference inside AWS, with no third-party AI provider in the path.
- Bundled
- Invoked in the deployment's AWS account.
- Tenant-account
- Invoked in the tenant's own AWS account, through a cross-account role. Spend bills to the tenant.
Never sold as local or on-prem: it is inference inside AWS. Add-on pricing to be announced.
Bring your own key
Separately from the tiers, bring your own provider keys and pay the provider directly.
- Your own LLM key
- OpenRouter is one option. A key changes who pays — never which models run.
- Jev key
- Same shape for decision judgments, independent of the vision-model choice.
- Bedrock role
- Billing identity and data boundary: with tenant-account Bedrock, images stay in the tenant's AWS account.
A key never changes how judgments are made — only who pays, and, for the Bedrock role, where inference lands.
Questions
When will pricing be announced?
Join the waitlist and we will email you when it opens.
Can I change tiers later?
Yes. Tell us what changed and we will move you; dedicated tiers are provisioned with you.
Do screenshots leave my systems?
They go to the vision model in memory and in transit only. We never store them, in any tier.
Is Bedrock local?
No. Bedrock is inference inside AWS, with no third-party AI provider in the path. Truly local inference means self-hosted open-weight models on your own hardware.