Deployment Mode Matrix
AxonFlow has both edition tiers and deployment modes. Teams often mix those concepts together, but they answer different questions:
- tier controls feature limits and licensing behavior
- deployment mode controls how the platform is operated and what the portal exposes
This page is the fast reference for the deployment side of that picture.
The Important Distinction
For public docs, the most relevant runtime and deployment states are:
- Community: self-hosted runtime with no license
- Evaluation: self-hosted runtime with an evaluation license
- SaaS: enterprise-operated multi-tenant deployment
- In-VPC: enterprise deployment in the customer’s AWS account
Community and Evaluation usually share the same core self-hosted topology. SaaS and In-VPC are enterprise operating models that change visibility, identity, and portal behavior.
Deployment Mode Matrix
| Mode | Typical user | Tenant isolation | Node visibility | Platform metrics visibility | Deployment controls | Protected portal workflows |
|---|---|---|---|---|---|---|
| Community | engineer validating locally | self-managed, no portal RLS surface | local only | self-managed via Prometheus/Grafana | Docker Compose / self-managed | No |
| Evaluation | platform team validating pre-prod | self-managed | self-managed | self-managed | self-managed | Usually no full enterprise portal path |
| SaaS | enterprise customer using AxonFlow-hosted control plane | yes | hidden in portal | tenant-focused usage only | hidden in portal | Yes |
| In-VPC | enterprise platform team running AxonFlow in its own AWS account | no RLS-style tenant isolation in portal mode | visible in portal | visible in portal | visible in portal | Yes |
What The Enterprise Portal Actually Uses
The customer portal’s deployment-config API currently returns:
modetenant_isolationshow_node_usageshow_platform_metricsshow_deployments
For the current portal code path:
saasmeans request-focused, tenant-isolated portal behaviorinvpcmeans platform-aware portal behavior with node and deployment visibility
That matters because the same enterprise portal intentionally behaves differently depending on where it is running.
How To Decide Which Mode Fits
Community
Choose community when you want to:
- stand up the core runtime quickly
- validate SDK integration
- test MCP, policies, and workflows locally
- prove that AxonFlow can govern real AI traffic before involving procurement
Evaluation
Choose evaluation when you need to:
- keep the same self-hosted control plane shape
- raise limits to something closer to real staging or pilot usage
- test approval queues, policy simulation, and evidence-export behavior
SaaS
Choose SaaS when the enterprise wants:
- portal-backed operations
- strong tenant isolation in the hosted control plane
- less infrastructure responsibility on the customer side
In-VPC
Choose In-VPC when the enterprise wants:
- AWS-account ownership
- more infrastructure visibility in the portal
- deployment and node operations inside its own environment
Common Mistake
The most common planning mistake is to treat Evaluation like a separate hosting mode. It is usually better understood as a licensed version of the self-hosted runtime. SaaS and In-VPC are the bigger operating-model changes.
Related Docs
Telemetry's deployment_mode Field
The deployment-mode concept above describes the operator's runtime topology. AxonFlow's anonymous heartbeat surfaces a closely related (but narrower) value on the wire as the deployment_mode field.
| Wire value | When SDKs / plugins set it |
|---|---|
community_saas | The configured AxonFlow endpoint host matches *.try.getaxonflow.com, or the env var AXONFLOW_TRY=1 is set |
self_hosted | Any other endpoint configuration |
unknown | Endpoint is empty or unparseable |
The wire enum is intentionally a closed three-value set so analytics queries don't have to enumerate every operator-side topology variant. The Pilot / Growth / Enterprise tier the operator sits in is not reflected on the wire — telemetry is anonymous and does not carry license-tier information. For the full wire-level field reference, see Telemetry and Checkpoint Service API.
