Skip to main content

Open-Source and AISIX Cloud Operating Models

AISIX is available as the open-source AISIX gateway or AISIX Cloud. Both use the same gateway runtime, which always runs in customer-managed infrastructure.

The open-source AISIX gateway is available under Apache 2.0 and runs as a standalone product without a control plane. AISIX Cloud is API7's commercial product for managing AISIX gateways. It adds control-plane workflows for resource management, gateway certificate issuance, usage reporting, and budgets.

AISIX Cloud has two control-plane deployment options:

  • Hybrid Cloud: API7 hosts the control plane. You run the gateway data plane in your infrastructure and manage resources through the dashboard or AISIX Cloud Admin API.
  • On-Premises: You host the control plane and run the gateway data plane in your infrastructure, including fully air-gapped environments. You manage resources through the same dashboard or AISIX Cloud Admin API.

Both deployment options provide the same resource model and gateway workflow. The difference is who hosts the control plane and where its data is stored.

Deployment Topology

The topology differs in where the management layer runs and how configuration reaches the gateway. Operators access an AISIX Cloud control plane through its dashboard or the AISIX Cloud Admin API.

Compare Operating Models

Open-source AISIX gatewayHybrid CloudOn-Premises
Gateway runtimeRuns in your infrastructureRuns in your runtime environmentRuns in your infrastructure
Management layerDeclarative resources.yaml fileAPI7-hosted AISIX Cloud control plane with dashboard and AISIX Cloud Admin APICustomer-hosted AISIX Cloud control plane with dashboard and AISIX Cloud Admin API
Configuration deliveryYou declare dynamic resources in a resources.yaml file, set process-wide settings in startup configuration, and reload the gateway to apply changesThe control plane projects environment resources to AISIX gatewaysThe on-premises control plane projects environment resources to AISIX gateways
Operational ownershipYou operate the gateway, its resources file, management integrations, and upgradesYou operate the gateway; API7 operates the control planeYou operate the complete private deployment with commercial delivery and support options
Network boundaryGateway configuration stays in systems you operate; upstream traffic and exported telemetry follow configured destinationsLive proxy traffic bypasses the control plane; the gateway connects outbound to the API7-operated control plane for management and telemetryControl-plane data and management services stay in your environment; upstream traffic and exported telemetry follow configured destinations
Best fitTeams that need the gateway runtime and can build or operate their own management workflowsTeams that want AISIX Cloud control-plane workflows without operating the control planeTeams with data-residency, network-isolation, compliance, or air-gap requirements

In every option, applications call the AISIX gateway rather than the management layer. The gateway then calls the configured upstream model, MCP server, or A2A agent.

Shared Gateway Capabilities

The AISIX gateway provides the traffic-processing capabilities used by all three operating models:

  • OpenAI-compatible APIs for inference, Realtime, Files, Batches, and Fine-Tuning
  • Anthropic-compatible Messages and Cohere-style rerank requests
  • Provider credentials and model aliases for hosted providers and OpenAI-compatible private endpoints
  • Multi-target routing, retries, fallback, semantic routing, and ensemble models
  • Caller API keys, model access controls, request and token rate limits, and concurrency limits
  • Guardrails, response caching, metrics, logs, and external observability exporters
  • MCP and A2A proxying, upstream authentication, access control, traffic controls, and observability

With the open-source gateway, declare dynamic resources in a resources.yaml file and set process-wide settings in startup configuration. You also operate gateway upgrades, credential-handling processes, and external reporting systems.

AISIX Cloud adds the control plane and dashboard without replacing the gateway traffic-processing layer.

AISIX Cloud Capabilities

The AISIX Cloud control plane adds organization-level workflows and operational services around the gateway runtime.

CapabilityOpen-source AISIX gatewayAISIX Cloud
Gateway resource managementDeclarative resources.yaml file for dynamic resources, startup configuration for process settings, and self-managed workflowsDashboard and AISIX Cloud Admin API with environment-scoped resource management
Organizations and environmentsBuild or integrate your own ownership modelOrganization ownership, isolated environments, members, teams, and roles
Caller API keysDeclare API-key resources and model allowlists in the resources fileAdd ownership, team bindings, lifecycle workflows, usage attribution, and visual management
Provider credentialsManage provider-key resources and their storage in your own environmentEnvelope-encrypted, write-only provider secrets, environment eligibility, dependency visibility, and in-place credential rotation
Data-plane operationsDeploy, secure, monitor, and upgrade gateway instances yourselfCertificate issuance, mTLS bootstrap, registration, heartbeat, configuration projection, and status visibility
Usage and costExport logs and metrics to systems you operateRequest logs and Usage views for request volume, tokens, spend, and cache savings
Model pricingIntegrate your own pricing and accounting systemAISIX Cloud model-price catalog, organization overrides, and per-request cost calculation
BudgetsNot available without the AISIX Cloud control planeOrganization, environment, caller API key, provider key, team, and member budgets with warning or blocking enforcement
Audit trailUse gateway logs and external audit systemsRequest logs plus an audit trail for control-plane resource and access changes
MCP and A2A managementDeclare servers, agents, authentication, and caller access in the resources fileDashboard and AISIX Cloud Admin API workflows for servers, agents, tool or agent grants, usage, and audit
Deployment and deliveryCommunity documentation and customer-managed deploymentHybrid Cloud or commercial On-Premises delivery using Docker Compose, Helm, or an offline package

Commercial capabilities can vary by offering and release. Contact API7 when you need to confirm a capability for a specific deployment or commercial agreement.

Open-Source AISIX Gateway

The open-source option provides direct control of the gateway and its surrounding infrastructure without requiring a commercial AISIX subscription. It suits teams that need the gateway runtime and plan to integrate it into an existing platform or build their own management workflows. Your team remains responsible for infrastructure and operating costs.

The proxy listener serves application traffic. Dynamic resources are declared in a resources.yaml file that the gateway loads at startup and reloads on SIGHUP. The private admin listener remains available for reads and health checks; the AISIX Admin API write path is deprecated in favor of the declarative file. You own the deployment topology, admin access, credentials, monitoring, backups, and upgrades.

Start with the Open-Source AISIX Gateway Quickstart, then review Production Deployment before serving production traffic. To manage several gateways from one place, use the AISIX Cloud control plane instead — see the On-Premises AISIX Cloud Quickstart.

Hybrid Cloud

In Hybrid Cloud, the gateway runs in your environment and connects to the API7-hosted AISIX Cloud control plane.

The local Admin API is not available. You manage resources through the dashboard or AISIX Cloud Admin API, and the control plane projects environment-scoped configuration to the gateway. The gateway connects with certificate-based mTLS, serves live AI traffic, reports telemetry, and performs AISIX Cloud budget checks through its control-plane connection.

Live proxy traffic does not pass through the control plane. Applications call the AISIX gateway, and the gateway calls upstream providers or agent endpoints directly.

See AISIX Cloud and Connect an AISIX Gateway for the commercial operating model.

On-Premises

On-Premises keeps the AISIX Cloud control plane, its data, and gateway management services inside infrastructure you control.

You deploy the control-plane API, data-plane manager, dashboard, and database in your infrastructure. You manage resources through that dashboard or AISIX Cloud Admin API. The on-premises package provides the same resource model and gateway workflow as Hybrid Cloud. It supports online installation with Docker Compose or Helm and offline installation from an air-gapped package.

See On-Premises Deployment for installation options and On-Premises Configuration for deployment settings.

Move between Operating Models

The proxy API and gateway runtime stay consistent across the operating models, so application integrations can continue using the same AISIX endpoint conventions and model aliases.

The management path does change. Open-source gateway deployments load dynamic resources from a resources.yaml file. Gateways connected to AISIX Cloud receive resources from the control plane and do not expose the local AISIX Admin API. Plan to recreate or migrate gateway resources when moving to an AISIX Cloud control plane; do not treat the change as an in-place switch of the startup configuration.