Agent Registry
AxonFlow supports both file-based agent definitions and an enterprise database-backed agent catalog. The managed catalog lets platform teams treat agent definitions as operational assets rather than only static YAML checked into a repo.
Use this page when you need to understand which agent operations are available in every deployment and which require the enterprise registry.
What Is Available
Community read APIs
The base runtime exposes:
GET /api/v1/agentsGET /api/v1/agents/{id}GET /api/v1/agents/domain/{domain}POST /api/v1/agents/validate
Those routes are primarily file-registry views.
Enterprise write and test APIs
The enterprise binary adds:
POST /api/v1/agentsPUT /api/v1/agents/{id}DELETE /api/v1/agents/{id}POST /api/v1/agents/{id}/activatePOST /api/v1/agents/{id}/deactivatePOST /api/v1/agents/{id}/testGET /api/v1/agents/{id}/versions
The practical model is:
- read-only registry access in all editions
- managed registry lifecycle in enterprise deployments
Authentication and Scope
Enterprise registry operations are organization-scoped. The handler accepts:
X-Org-ID- tenant context derived from Basic auth credentials as a compatibility fallback
X-User-IDfor change attribution
If X-Org-ID is missing, write operations return an authorization error. For enterprise customers using the protected portal, this context is usually injected by the portal session and proxy layer.
List and Filter Agents
GET /api/v1/agents?page=1&page_size=20&domain=travel&search=booking&sort_by=updated_at&sort_dir=desc
Supported query parameters:
| Parameter | Purpose |
|---|---|
page | 1-based page index |
page_size | page size, capped at 100 |
domain | Filter to one domain |
search | Free-text filter |
sort_by | name, domain, created_at, or updated_at |
sort_dir | asc or desc |
is_active | Filter active or inactive agents |
The enterprise response follows the familiar list shape:
{
"agents": [],
"pagination": {
"page": 1,
"page_size": 20,
"total_items": 0,
"total_pages": 0
}
}
Create and Update Requests
Create requests use this shape:
{
"name": "travel-coordinator",
"domain": "travel",
"description": "Coordinates travel search and approval flows",
"config": {
"type": "coordinator"
},
"is_active": true
}
Update requests are partial:
{
"description": "Coordinates regulated travel and approval flows",
"is_active": false
}
That partial-update model is useful when platform teams want to change activation state or metadata without resubmitting the whole agent specification.
Activation, Deactivation, and Testing
Activate or deactivate
POST /api/v1/agents/{id}/activate
POST /api/v1/agents/{id}/deactivate
Use these routes when you want to stage an agent definition before exposing it, pause an unstable agent without deleting history, or switch traffic between old and new registry entries.
Sandbox test
POST /api/v1/agents/{id}/test
Request body:
{
"query": "Book a business trip to London next week",
"context": {
"department": "finance"
}
}
This returns a test result with whether the run succeeded, the matched rule or agent, the result payload when available, execution duration, and any error message.
It is not a full production canary, but it is a useful pre-check before enabling an agent in a live environment.
Version History
Every enterprise registry change is captured in agent-version records:
GET /api/v1/agents/{id}/versions
Version entries include:
- the agent snapshot
- version number
- change type such as
create,update,delete,activate, ordeactivate - who changed it
- when it changed
For regulated or high-accountability environments, this is one of the most important differences from file-only workflows.
How It Fits With File-Based Agents
Enterprise does not remove the file-based registry. It adds a managed database-backed layer on top of it.
Teams often use:
- file-based agents for bootstrap or shared defaults
- database-backed agents for tenant-specific or organization-specific operations
This hybrid model is often the right answer during migration from prototype to production.
Rollout Pattern
- Validate the new configuration with
POST /api/v1/agents/validate. - Create the enterprise registry entry.
- Run
POST /api/v1/agents/{id}/testagainst representative queries. - Activate the agent only after policy and approval paths are configured.
- Keep the version history link in incident and audit runbooks.
