Community vs Evaluation vs Enterprise

AxonFlow currently exposes three runtime capability tiers in the codebase:

• Community: self-hosted local/dev usage, no license required
• Evaluation: free evaluation license with elevated limits and selected approval/simulation features
• Enterprise: paid production tier for organization-wide rollout and advanced operational controls

This page focuses on what the runtime and APIs actually gate today.

Terminology note

The license/status APIs also expose paid SKU names such as Professional, Enterprise, and Plus. In the open docs, those all map to the broader Enterprise capability bucket because they share the enterprise code path and enterprise limit profile.

Quick Guidance

If you need to...	Best fit
Run AxonFlow locally with Docker Compose	Community
Add governance to a dev or staging system without buying production licenses yet	Evaluation
Test HITL queues, policy simulation, or evidence export before procurement	Evaluation
Roll out org-wide policies, portal workflows, or enterprise operations in production	Enterprise
Use production identity, protected portal workflows, or advanced compliance operations	Enterprise

The Runtime Differences That Matter

The tier system currently controls:

• custom tenant and organization policy limits
• maximum configured LLM providers
• execution history and concurrency limits
• SSE connection limits
• cost-estimation limits
• whether HITL approval, policy simulation, and evidence export are enabled

Capability Summary

Capability	Community	Evaluation	Enterprise
Local self-hosted runtime	Yes	Yes	Yes
Proxy mode and gateway mode	Yes	Yes	Yes
Workflow Control Plane	Yes	Yes	Yes
Tenant custom policies	Limited	Higher limit	Unlimited
Organization-tier policies	No	Yes	Yes
Configured LLM providers	Limited	Higher limit	Unlimited
HITL approval workflows	No queue	Enabled with limits	Enabled, production-ready
Policy simulation	No	Enabled with limits	Enabled
Evidence export	No	Enabled with limits	Enabled
Portal, SCIM, SSO, protected enterprise docs	No	No	Yes

Current Limit Profiles

These values come directly from the non-enterprise license tier definitions used by the runtime.

Limit	Community	Evaluation	Enterprise
Tenant policies	20	50	Unlimited
Organization policies	0	5	Unlimited
Connectors with custom policies	2	5	Unlimited
LLM providers	2	3	Unlimited
Audit retention	3 days	14 days	Configurable, up to long-term retention
Execution history	50	500	Unlimited
Concurrent executions	5	25	Unlimited
Plans	25	100	Unlimited
Versions per plan	10	25	Unlimited
SSE connections	5	25	Unlimited
Cost estimates per day	10	100	Unlimited
Pending approvals	5	25	Unlimited

What Evaluation Adds

Evaluation is the “try the real workflows” tier. It is the first tier where the runtime enables:

• organization-tier policies
• queue-backed approval behavior for WCP require_approval
• session overrides — create time-bounded, audit-logged policy overrides (60m default, 24h cap)
• workflow checkpoint resume — resume an interrupted workflow from its last checkpoint
• extended decision explainability retention (30 days vs 7 in Community)
• policy simulation
• evidence export
• higher limits for plans, executions, and provider configuration

In the current runtime:

• Evaluation approvals have a fixed 24-hour expiry
• Evaluation evidence export and simulation are still limited
• Evaluation keeps the same public API surface as Community, but more features actually activate

The practical evaluation case

Evaluation is what most teams need once AxonFlow is being reviewed by more than one enthusiastic engineer.

It is usually the right next step when:

• a platform team wants realistic validation, not just a local demo
• security or compliance stakeholders want to see approval or evidence workflows
• you need more headroom before committing to a production operating model

Request a license here:

• Request an Evaluation License
• Evaluation Rollout Guide
• Assessing AxonFlow in Regulated Environments

What Enterprise Adds

Enterprise is about production operations, protected configuration surfaces, and organizational controls.

That includes:

• protected enterprise documentation
• enterprise portal workflows and admin surfaces
• SCIM, SSO, and enterprise identity features
• enterprise-only connector and provider operations
• advanced compliance and operational playbooks
• org-wide override administration (listing and revocation across tenants)
• auto-approval of low-risk HITL requests and SLA escalation of critical ones
• resume-from-any-checkpoint for workflow recovery, not just resume-from-last
• unlimited or effectively unbounded runtime limits in the current tier model

The practical enterprise case

Enterprise is what organizations usually need when AxonFlow moves from “useful in one workflow” to “part of how we govern AI across the company.”

At that point, the differentiator is not only more headroom. It is the combination of:

• organization-wide governance controls
• protected admin and portal workflows
• identity and provisioning
• enterprise-grade connector and provider operations
• stronger audit, oversight, and compliance posture
• support and deployment expectations that fit business-critical usage

Connectors by Edition

The connector implementation split in the codebase is clearer than the current docs structure:

Core connector types available in the base/community factory

• postgres
• mysql
• mongodb
• cassandra
• redis
• http
• s3
• azure_blob
• gcs

Connector types added by enterprise build paths

• amadeus
• salesforce
• slack
• snowflake
• hubspot
• jira
• servicenow

That means connector availability is more nuanced than “all public connector docs are community-safe.” If you are planning around SaaS-business-system connectors, verify the target deployment tier before committing.

LLM Providers by Edition

Built-in provider types in the runtime:

• openai
• anthropic
• azure-openai
• gemini
• ollama
• custom
• bedrock (Enterprise only)

Practical edition guidance:

• Community and Evaluation are sufficient for most local and staging provider work
• Enterprise is the right assumption for Bedrock-centric, managed-credential, or production multi-provider operations

Approval Behavior Differences

The approval story is mode- and tier-sensitive:

• In gateway mode, Community does not provide a real HITL queue
• In WCP, Community can reach require_approval policy decisions, but queue-backed approval only activates in Evaluation or Enterprise
• Evaluation enables approval queues with fixed expiry and limit enforcement
• Enterprise removes the evaluation-tier operational limits

Simple Rule Of Thumb

Use this mental model:

• Community is for getting AxonFlow into engineers’ hands quickly.
• Evaluation is for proving AxonFlow can survive the first serious production-readiness conversation.
• Enterprise is for running AxonFlow as a scalable control layer with stronger guarantees across the organization.

Recommended Upgrade Path

1. Start in Community to wire AxonFlow into your app and validate the developer experience.
2. Move to Evaluation when you need real approval queues, simulation, and larger plan/provider limits.
3. Move to Enterprise when identity, portal workflows, compliance operations, or production operational control become requirements.

Full Feature Matrix

The matrix below is the complete reference — every governance, runtime, identity, and operations feature with its tier-by-tier availability. Use it for procurement reviews, security reviews, or anytime the headline summary above isn't specific enough. Sections are ordered from most commonly checked at the top (Governance, Runtime) to operational depth at the bottom (Compliance, Support, Roadmap).

Show / hide full matrix

1. Governance & Security

Feature	Community	Evaluation	Enterprise
Policy Enforcement
Policy enforcement engine	✅	✅	✅
Sub-10ms inline governance	✅	✅	✅
System Policy API (list, get, CRUD)	✅	✅	✅
Tenant Policy CRUD API	✅	✅	✅
Pattern testing API	✅	✅	✅
Policy risk_level + allow_override metadata	✅	✅	✅
Session overrides (time-bounded, audit-logged)	Context only	Create, 60m default / 24h cap	+ org-wide listing, admin revoke
Decision explainability endpoint	✅ 7-day retention	✅ 30-day retention	✅ 365-day, org-wide
Threat Detection
PII detection (SSN, credit cards, PAN, Aadhaar)	✅	✅	✅
SQLi response scanning (basic)	✅	✅	✅
Advanced SQLi detection (ML-assisted)	❌	❌	✅
Code Governance
Code artifact detection in LLM responses	✅	✅	✅
Language detection (14 languages)	✅	✅	✅
Code type categorization	✅	✅	✅
Secret pattern detection	✅	✅	✅
Unsafe code pattern detection	✅	✅	✅
Git provider integration (GitHub, GitLab, Bitbucket)	❌	❌	✅
PR creation from LLM-generated code	❌	❌	✅
Code governance dashboard	❌	❌	✅
Media Governance
Image validation (format, size, dimensions)	✅	✅	✅
OCR-based PII detection	✅	✅	✅
SHA-256 audit hashing	✅	✅	✅
System media policies (5 default rules)	When enabled	✅	✅
Toggle system media policies	✅	✅	✅
Modify system media policy actions/priority	❌	❌	✅
Per-tenant media governance config	❌	❌	✅
Cloud analyzers (Rekognition, Vision, Azure)	❌	❌	✅
Identity & Access
SSO/SAML integration	❌	❌	✅
SCIM 2.0 user provisioning	❌	❌	✅
SCIM group sync & role mapping	❌	❌	✅
Evaluation & Simulation
HITL Approval Gates	No queue	Max 100 pending, 24h expiry	Unlimited, configurable
Risk-tiered approval routing	Severity metadata only (no queue)	Severity + queue filter	+ auto-approve low-risk, SLA escalation for critical
Policy Simulation (dry-run)	❌	300/day	Unlimited
Impact Report	❌	50 inputs/run	100 inputs/run
Evidence Export Pack	❌	14-day window, 5K records, 3/day, watermark	Unlimited, clean

2. Audit & Compliance

Feature	Community	Evaluation	Enterprise
Audit Logging
Audit logging (LLM calls)	✅	✅	✅
MCP query audit logging	✅	✅	✅
Decision chain tracing	✅	✅	✅
Transparency headers (X-AI-*)	✅	✅	✅
Audit retention	3 days	14 days	Configurable (up to 10 years)
EU AI Act
Conformity assessment workflow	❌	❌	✅
10-year audit retention	❌	❌	✅
EU AI Act export format	❌	❌	✅
India Compliance (SEBI/RBI)
India PII detection (Aadhaar, PAN, UPI)	✅ Pattern-based	✅ Pattern-based	✅ With checksum
SEBI compliance module (export, 10-year audit retention)	❌	❌	✅
RBI FREE-AI Framework (kill switch, board reports)	❌	❌	✅
Singapore Compliance (MAS FEAT)
Singapore PII detection (NRIC, FIN, UEN)	✅ Pattern-based	✅ Pattern-based	✅ With checksum
FEAT assessment workflows	❌	❌	✅
MAS-compliant export format	❌	❌	✅

3. Policy Management

Feature	Community	Evaluation	Enterprise
83 System policies (SQLi, PII, compliance, code, security)	✅ View	✅ View	✅ View + Override
Tenant-tier policies	20 limit	50 limit	Unlimited
Organization-tier policies	❌	5 limit	Unlimited
System policy overrides (block/warn/log)	❌	❌	✅
Policy version history	Last 5	Last 5	Full audit trail
require_approval action	No queue	Queue with 24h expiry	Full HITL queue
Policy templates library	Core	Core	Full (EU AI Act, SEBI, RBI, MAS FEAT)
Customer Portal Policy UI	❌	❌	✅

4. LLM Providers & Routing

Feature	Community	Evaluation	Enterprise
Providers
OpenAI	✅	✅	✅
Azure OpenAI	✅	✅	✅
Anthropic (Claude)	✅	✅	✅
Google Gemini	✅	✅	✅
Ollama	✅	✅	✅
AWS Bedrock	❌	❌	✅
Routing
Multi-provider failover	✅	✅	✅
Weighted routing	✅	✅	✅
Round-robin routing	✅	✅	✅
Health-based selection	✅	✅	✅
Cost-optimized routing	❌	❌	✅
Runtime weight updates	❌	❌	✅
Cost Management
Usage tracking (tokens, cost)	✅	✅	✅
Budget limits and alerts	✅	✅	✅
Pre-flight budget check	✅	✅	✅
Cost estimation	10/day	100/day	Unlimited
Usage forecast	❌	❌	✅
Cost dashboard	❌	❌	✅

5. MCP Connectors & Data Access

Feature	Community	Evaluation	Enterprise
Community Connectors
PostgreSQL, MySQL, MongoDB	✅	✅	✅
Redis, HTTP/REST, Cassandra	✅	✅	✅
S3, Azure Blob, GCS	✅	✅	✅
Enterprise Connectors
Amadeus, Salesforce, Slack	❌	❌	✅
Snowflake, HubSpot, Jira, ServiceNow	❌	❌	✅
MCP Policy Enforcement
SQL injection blocking (INPUT phase)	✅	✅	✅
PII redaction (OUTPUT phase)	✅	✅	✅
Exfiltration detection	✅	✅	✅
Connectors with custom policies	2	5	Unlimited

What Evaluation Adds

Feature	Community	Evaluation	Enterprise
Deployment
Docker Compose (local)	✅	✅	✅
AWS ECS/Fargate	Manual	Manual	One-click CloudFormation
Multi-tenant isolation	❌	❌	✅
Workflow Control Plane
Workflow registration (LangChain, LangGraph, CrewAI)	✅	✅	✅
Step gates (allow/block)	✅	✅	✅
Per-tool governance	✅	✅	✅
LangGraph adapter (Python, TypeScript, Go, Java)	✅	✅	✅
MCP tool interceptor	✅	✅	✅
SDK support (Go, Python, TypeScript, Java)	✅	✅	✅
Per-step token/cost tracking	✅	✅	✅
SSE real-time execution streaming	✅	✅	✅
Idempotent step-gate retry (execution boundary semantics)	✅	✅	✅
Retry context on step gates (wire-level — gate_count, prior_completion_status, prior_output)	✅	✅	✅
Idempotency key on step gates (wire-level, same-workflow enforcement)	✅	✅	✅
Retry-aware policy conditions (step.gate_count, step.prior_completion_status, step.idempotency_key, etc. as policy conditions)	❌	✅	✅
Workflow checkpoint listing	✅	✅	✅
Resume from checkpoint	❌	From last	From any
Execution Timeline UI	❌	❌	✅
Circuit Breaker
Manual trip/reset API	✅	✅	✅
Auto-trip (sliding window, error/violation counting)	❌	❌	✅
Per-tenant circuit breaker config	❌	❌	✅
Circuit breaker notifications (webhook, Slack, PagerDuty)	❌	❌	✅
HITL & Approvals
HITL Queue API	❌	✅ Limited	✅
HITL Approval Dashboard UI	❌	❌	✅
WCP plane-scoped approve/reject (/workflows/{id}/steps/{step_id}/approve|reject)	❌	✅	✅
MAP plane-scoped approve/reject (/plans/{id}/steps/{step_id}/approve|reject)	❌	✅	✅
WCP-plane pending listing (GET /api/v1/workflows/approvals/pending)	❌	✅	✅
MAP-plane pending listing with plan_id + ?plan_id= filter (GET /api/v1/plans/approvals/pending)	❌	✅	✅
MAP confirm / step execution mode (plan-level HITL)	❌	❌	✅
Cross-plane response parity (same retry_context, approval_id, policies_matched, plane-scoped pending endpoints)	❌	✅	✅
Support
Community support (GitHub Issues)	✅	✅	✅
Priority support & SLA	❌	❌	✅
Customer Portal UI	❌	❌	✅

Related Docs

• Getting Started
• Choosing a Mode
• LLM Providers Overview
• MCP Connectors Overview
• Enterprise Overview
• When Community Stops Being Enough
• Evaluation Rollout Guide