Worker crash
Execution state persists. Tasks are retried.
Multi-region & multi-cloud resilience
Agentic workflows.
Resilient across regions and clouds.
Orkes Conductor orchestrates your AI agents, workflows and services with durable execution, monitors cluster health, and fails over automatically.
99.99% Uptime SLA for
paired deployments
paired deployments
Orkes-hosted or customer-hosted. Your cloud accounts or ours.
Extending the reliability you already trust
Durable execution, beyond a single region
Orkes already protects execution through worker failures and availability-zone outages. Multi-region resilience adds recovery in another region or cloud.
Zone outage
Multi-AZ deployment protects the cluster.
Worker crash
Region outage
A replica in another region is ready to take over.
Zone outage
Worker crash
Cloud outage
Execution can resume in another cloud.
Region outage
Zone outage
Worker crash
Managed recovery
Managed recovery
Orkes monitors cluster health continuously and automatically manages failover. Your agentic workflow resumes on the second cluster.
Illustrative multi-cloud deployment
One multi-region endpoint Routes requests to Region A
Region A Running Primary cluster
Serves traffic and runs your agentic workflows.
Services · AI agents · Approvals
Region B Running Replica cluster
Continuously receives state. Ready to take over.
Execution state + configuration
10 min Recovery target
90 sec Maximum data-loss target for recent execution progress
~0 sec Target RPO for workflows in synchronous mode
Recovery can be much faster. These are recovery objectives, not guarantees.
Deployment choice
Your cloud accounts or ours
Run across AWS, Microsoft Azure and Google Cloud, in Orkes-hosted or customer-hosted deployments. Orkes manages replication and failover in either model.
Multi-region
Pair clusters in two regions of the same cloud provider.
Region A Primary
Region B Replica
Multi-cloud
Pair clusters across cloud providers for recovery beyond a single cloud.
AWS
Primary · Region A
Google Cloud
Replica · Region B
Built for critical workloads
Security and performance
Encryption in transit and at rest
Replication is encrypted end to end between the clusters. Data is encrypted at rest in both locations, in the regions and cloud accounts selected for your deployment.
Capacity for replication and recovery
Multi-region resilience is designed to preserve core workflow performance. Orkes allocates the capacity required for replication and recovery.
Further detail
Common questions
What happens to work already running?
The agentic workflow resumes on the second cluster using its copy of the execution state. Some in-flight tasks may be retried. Business actions, such as issuing a payment, should handle retries without duplicating those actions.
How do the two clusters work together?
Both clusters are running and monitored by Orkes. The primary serves traffic and executes workflows. The replica continuously receives execution state and configuration, ready to take over automatically. Orkes also restores an unhealthy replica to health.
What gets replicated?
Running execution state, workflow and task definitions, prompts, schedules, secrets, integrations, users and access configuration. Execution history and audit logs are replicated from the time the clusters are paired.
Will applications need a new address after failover?
Applications and workers using Orkes’s multi-region endpoint keep the same address. Orkes updates routing to the active cluster.
What do I need to keep available outside Orkes?
Your workers and external dependencies need their own availability arrangements. Workers must be able to reach the active Orkes cluster, and services such as databases and model providers must remain accessible to your application.
What does the uptime SLA apply to?
The 99.99% uptime SLA applies to the paired deployment. A single multi-AZ cluster provides a 99.9% uptime SLA. Recovery time and data-loss objectives are separate targets.
How is multi-region resilience priced?
The capability is available as an add-on for enterprise deployments. Contact Orkes for pricing and further information. See enterprise plans.
Build on Orkes
Build agentic workflows that connect AI agents, business services and human decisions. Run your agents and business workflows on the same durable execution platform.