Skip to main content

Deploy AxonFlow

AxonFlow Community is designed to run self-hosted. The default local deployment uses Docker Compose and brings up the same core runtimes engineers use in trial and early production:

  • Agent on :8080 for inline policy enforcement and MCP access
  • Orchestrator on :8081 for workflow execution, routing, and multi-agent control
  • PostgreSQL for platform state and audit data
  • Redis for runtime coordination and rate limiting
  • Prometheus and Grafana for observability

Community is the fastest path to understand the platform end to end. Evaluation and Enterprise add higher limits, identity, compliance, and enterprise deployment workflows once a team is moving from pilot to production.

This page is intentionally community-first. If you are an engineer trying to understand whether AxonFlow can run as the governed control plane for your AI stack, this is the place to start. If you already know you need AWS-native enterprise rollout paths, use this page to understand the runtime shape first, then move into the AWS Marketplace and CloudFormation deployment guides.

Deployment Options​

OptionDescriptionBest For
Self-HostedDocker Compose deployment for local, trial, and smaller self-managed environmentsEngineers validating the platform quickly
AWS MarketplaceAWS Marketplace deployment path backed by CloudFormationEnterprise rollout and procurement workflows
CloudFormationAWS infrastructure-as-code deploymentEnterprise teams with custom VPC, networking, and controls

How Most Teams Progress​

The typical journey looks like this:

  1. start with Community Docker Compose to validate SDK integration, policies, MCP, and workflows
  2. use Evaluation when the team needs higher limits and more production-like governance features
  3. move to Enterprise when procurement, identity, compliance, and enterprise deployment workflows become part of the rollout

That progression is useful because it mirrors how serious AI products are usually adopted: first by engineers, then by platform teams, then by broader enterprise stakeholders.

Run It​

The runnable steps live in two places, and this page does not repeat them:

  • Getting Started takes you from git clone to a running local stack and your first governed request, with the local system requirements.
  • Self-Hosted Deployment is the reference for that same Docker Compose stack: every service and port (including the local Grafana login), the verification commands, the configuration variables that matter first, and hardening before shared use.

Two things worth knowing before you verify: the Agent (:8080) and Orchestrator (:8081) /health endpoints return JSON with status, service, version and capability metadata, and Prometheus scrapes /prometheus. /metrics is a JSON endpoint kept for platform and debug flows, not a scrape target.

Readiness Checklist​

  • Add at least one LLM provider key if you need proxy mode, routed workflows, or MAP
  • Review LLM Providers and Choosing a Mode
  • Confirm MCP connector configuration if your workflow needs governed database or API access
  • Verify Prometheus and Grafana so you can observe latency, blocked requests, and token/cost activity from day one
  • Use Deployment Mode Matrix and Capacity Planning And Sizing before you commit to a larger pilot or shared environment

What A Staff Engineer Usually Wants To Prove​

Before recommending AxonFlow for broader adoption, a senior or staff engineer usually wants to show:

  • the local stack is easy to run repeatedly
  • the request path is observable
  • policies behave predictably
  • the platform can support the multi-agent or connector-heavy workflows the team actually plans to build

That is why the deployment story should not stop at "containers started." It should end at "we ran a realistic governed workflow and know how it behaved."

Sizing A Shared Environment​

The minimums for a local trial are in Getting Started. Once more than one engineer depends on the stack, plan for:

  • 4+ vCPU
  • 8-16 GB RAM
  • Persistent PostgreSQL storage
  • Centralized logs plus Prometheus/Grafana retained outside a laptop

Production Direction​

Community is enough to build and validate sophisticated governed AI systems. When teams need larger limits, enterprise identity, procurement-friendly deployment, and stronger governance guarantees, the natural next step is Evaluation or Enterprise, along the progression described above.

If you are already mapping that journey, use these pages together:

Next Steps​

Operational Readiness Checklist​

Before relying on this page in a production rollout, pair it with the core operations docs:

Last updated on