Choosing an AI agent deployment strategy involves more than selecting a model or connecting workplace applications. IT teams must evaluate infrastructure ownership, access controls, network policies, maintenance requirements, and the operational effort needed to keep the system running.
Omni gives organizations three deployment paths: Docker Compose, Terraform on AWS or GCP, and Omni Cloud, which is coming soon. Each option serves a different operational preference, from running an instance on infrastructure your organization controls to choosing a managed service when it becomes available.
This guide compares these options and explains what IT, infrastructure, and security teams should evaluate before making a deployment decision.
Three AI agent deployment options
The right AI agent deployment approach depends on your existing infrastructure, operational requirements, security policies, and preferred level of control. Organizations can begin with a straightforward deployment for evaluation or use infrastructure as code workflows for production cloud environments.
Omni offers three deployment options, each suited to a different operational approach.
Scroll the table horizontally to compare all columns.
| Deployment option | Description | Best suited for |
|---|---|---|
| Docker Compose | Run Omni on infrastructure controlled by your organization using Docker Compose. | Product evaluation and smaller deployments. |
| Terraform on AWS/GCP | Deploy Omni in AWS or GCP using the official Terraform modules and infrastructure as code workflows. | Organizations deploying Omni in production cloud environments. |
| Omni Cloud | A planned managed deployment option that lets teams use Omni without managing the underlying infrastructure themselves. | Organizations that prefer a managed service. Omni Cloud is coming soon and is not generally available. |
Paid Business plans apply regardless of which deployment path the customer chooses; Community remains free to self-host.
These options differ mainly in how teams deploy and operate the software. Docker Compose offers a straightforward route to a running instance, while Terraform supports infrastructure as code workflows in AWS and GCP. Omni Cloud is intended for organizations that prefer a managed service instead of operating the deployment themselves.
Before choosing an option, consider who will maintain the environment, how your organization manages infrastructure changes, and which network and data controls your policies require.
Find service settings and environment variables in our configuration reference.
Docker Compose: the fastest path to a running instance
Docker Compose provides a straightforward way to run Omni on infrastructure your organization controls. It is suitable for teams evaluating the product and organizations planning smaller deployments without starting with a production cloud infrastructure workflow.
This option allows IT teams to assess the product in their own environment and understand how it fits their existing systems before deciding on a longer term deployment strategy. The Omni CLI supports upgrades and diagnostics, helping teams manage and troubleshoot their instances. Your team remains responsible for the infrastructure and its ongoing operation.
Follow our Docker Compose guide for installation, configuration, and hardware requirements.
Terraform for production on AWS and GCP
We provide Terraform modules for production deployments on Amazon Web Services (AWS) and Google Cloud Platform (GCP).
Terraform enables infrastructure as code workflows. Instead of managing infrastructure changes only through manual processes, teams can use configuration files to describe and manage their deployment. This approach can support repeatability, change reviews, and alignment with established infrastructure management practices.
To deploy on AWS, follow our AWS Terraform guide. For GCP, use our Terraform modules.
Choose AWS or GCP based on your existing cloud environment, internal policies, and operational capabilities. Our deployment overview explains the available paths and what your team operates.
Terraform does not remove the need for operational oversight. Your team must still understand the documented configuration, review changes, and maintain the resulting environment according to its requirements.
Our AWS Terraform guide walks through the AWS deployment. For GCP, use our Terraform modules and contact us for rollout guidance.
Isolated environments and on premise deployment
Some organizations have strict requirements about where company information is processed and which systems can communicate with external services. An on premise AI agent deployment can be relevant when internal infrastructure policies require greater control over the environment.
Omni can run in your organization's cloud, on premises, or in an isolated environment. These deployment choices are relevant for teams that need to control infrastructure, restrict network access, or keep application operations within an approved environment.
Omni also supports self hosted model endpoints such as vLLM and Ollama. This gives organizations the option to connect compatible model endpoints running within their own environment rather than relying exclusively on external model services.
When the deployment and model endpoints remain internal, organizations can design an AI operation around their network boundaries and data handling policies. However, the actual level of isolation depends on the deployment configuration, connected services, model endpoints, and network rules. Teams should verify these details before treating an environment as fully isolated.
Infrastructure control is only one part of security. IT teams must also consider which employees can access company information through the agent.
Our permission-aware company context applies the permission metadata supplied by each connector when retrieving information. Visibility rules vary by source. See our connector access details before connecting sensitive information.
Configure services, model providers, and authentication with our configuration reference.
Omni Cloud: managed, when you want it
Not every organization wants to maintain the infrastructure behind an AI agent. Some teams prefer to focus on adoption, integrations, and internal use cases while a provider manages the underlying service.
Omni Cloud is intended for organizations that prefer a managed deployment rather than operating the infrastructure themselves. The planned service is designed to provide a managed alternative to running Omni within an organization's own environment.
Omni Cloud is not currently generally available. Organizations interested in the managed option can visit the Omni Cloud waitlist to receive updates when the service becomes available.
IT teams should account for this status when planning implementation timelines. The waitlist does not mean that customers can deploy through Omni Cloud today.
Organizations that need to evaluate the product now can explore the documented Docker Compose deployment path or review the Terraform options for AWS and GCP. These routes allow teams to assess the software within an environment they control while they continue to evaluate their longer term infrastructure requirements.
What about hardware requirements?
Infrastructure requirements depend on the deployment setup. A single set of hardware specifications would not necessarily apply to every organization because deployments can differ in scale, connected systems, data volume, and model configuration.
When assessing requirements, IT teams should consider four factors:
- Deployment shape: The selected deployment approach and how the organization intends to operate the instance.
- Connector count: The number and types of connected workplace systems that Omni needs to work with.
- Index size: The volume of information that must be indexed and made available for retrieval.
- Model setup: Whether the organization uses external model services or compatible endpoints within its own environment.
These factors can influence resource planning and operational needs. Teams should evaluate them against their expected workload rather than assume that a configuration suitable for a small evaluation will also suit a larger deployment.
We keep hardware requirements in our Docker Compose guide and deployment settings in our configuration reference so you can use the right specifications for your setup.
Before deploying in production, validate the configuration against your workloads, policies, and operating environment.
FAQs
Can I try Omni before committing to a full deployment?
Yes. You can evaluate our free Community edition on your own infrastructure with Docker Compose. Follow our Docker Compose guide to set up an instance and connect your tools. You cover the infrastructure and model usage costs.
Is Omni Cloud available today?
No. Omni Cloud is our planned managed service and is not yet generally available. Join our waitlist to get notified when it launches. You can evaluate Omni today with Docker Compose or a self-managed Terraform deployment.
Does self hosting mean we maintain everything ourselves?
Your team operates the infrastructure, including monitoring, backups, access controls, and security. Our CLI helps with upgrades and diagnostics for Docker Compose deployments. Business includes priority email support and guided onboarding; you can use it while continuing to self-host.
Which cloud providers are supported?
We provide Terraform modules for AWS and GCP. Our step-by-step deployment guide covers AWS; GCP setup uses the modules in our repository. Contact us for help planning a GCP rollout. Docker Compose also works on infrastructure that meets its prerequisites.