Agent Hosting

Deploy persistent AI agents on sovereign GPU nodes. Agents get inference access, knowledge base access, persistent memory, health monitoring, and hourly billing.

Deploy an agent

curl -X POST https://api.ryvion.ai/api/v1/agents/deploy \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "research-agent",
    "container_image": "ghcr.io/myorg/research-agent:latest",
    "kb_ids": "kb_abc,kb_def",
    "model": "phi-4",
    "jurisdiction": "CA"
  }'

Request body

Parameter	Type	Required	Description
`name`	string	Yes	Agent name
`container_image`	string	Yes	OCI container image URL
`kb_ids`	string	No	Comma-separated knowledge base IDs the agent can access
`model`	string	No	Default model for inference (e.g., `phi-4`)
`jurisdiction`	string	No	Jurisdiction to deploy in (e.g., `CA`, `DE`)

What your agent gets

When deployed, your agent container runs on a GPU node with:

Inference access -- call the Ryvion API from inside the container using the assigned model
Knowledge base access -- read and search the knowledge bases specified in kb_ids
Persistent memory -- store and recall memories through the Agent Memory API
Health monitoring -- the hub monitors agent health and restarts on failure
Hourly billing -- billed at $0.50 CAD/hour while running

List deployments

curl https://api.ryvion.ai/api/v1/agents/deployments \
  -H "Authorization: Bearer YOUR_API_KEY"

Response format

{
  "deployments": [
    {
      "id": "deploy_abc123",
      "name": "research-agent",
      "status": "running",
      "node_id": "node_xyz",
      "jurisdiction": "CA",
      "created_at": "2026-04-08T12:00:00Z"
    }
  ]
}

Stop an agent

curl -X POST https://api.ryvion.ai/api/v1/agents/deployments/deploy_abc123/stop \
  -H "Authorization: Bearer YOUR_API_KEY"

Endpoints summary

Method	Endpoint	Description
`POST`	`/api/v1/agents/deploy`	Deploy an agent
`GET`	`/api/v1/agents/deployments`	List all deployments
`POST`	`/api/v1/agents/deployments/:id/stop`	Stop a running agent

Building your agent container

Your container image should:

Accept environment variables for API credentials and configuration
Use the Ryvion API (or OpenAI SDK pointed at the hub) for inference
Optionally use the knowledge base and memory APIs for state
Implement a health check endpoint if you want monitoring

Jurisdiction pinning

Specify a jurisdiction to ensure your agent runs on a node in a specific country. See Sovereign Routing for available jurisdictions.

Pricing

$0.50 CAD per hour while the agent is running. Billing stops when the agent is stopped or terminated. Inference, search, and memory calls made by the agent are billed separately at standard rates.