POST /mltk/agents

Description. Create a new agent. The handler:

1. Inserts a record into aitk_agent_collection with state "Creating".
2. Calls SCS POST {scs_url_base}/agent/aws_agentcore to provision the runtime.
3. Persists the returned runtime_id and version.
4. Dispatches an asynchronous agentstatus search to monitor convergence.

Body (JSON).

Field	Required	Type	Default	Notes
name	yes	string	—	Unique agent name
runtime_type	no	string	"AWS_AGENT_CORE"	Runtime platform
system_prompt	no	string	"You are a helpful assistant."
task_prompt	no	string	""
description	no	string	""
llm	no	object	{}	{ "provider": ..., "model": ... }
mcps	no	array	[]	List of { "name": ..., "tools": [...] }
knowledge_bases	no	array	[]	Vector store references
agent_timeout	no	int	450	Seconds
acl	no	object	owner-share	{ "sharing": ..., "perms": { "read": [], "write": [] } }

Response (200).

{
  "agent_name": "demo",
  "runtime_type": "AWS_AGENT_CORE",
  "tracking_job_id": "1715425200.123"
}

Example.

curl -sk -u "$SPLUNK_USER:$SPLUNK_PASSWORD" -X POST 
  -H "Content-Type: application/json" 
  --data @- "$SPLUNK_HOST/servicesNS/nobody/Splunk_ML_Toolkit/mltk/agents?output_mode=json" <<'JSON'
{
  "name": "demo",
  "description": "Quick demo agent",
  "system_prompt": "You are a helpful Splunk admin.",
  "task_prompt": "Answer with SPL examples.",
  "llm":  { "provider": "OpenAI", "model": "gpt-4o-mini" },
  "mcps": [ { "name": "splunk_mcp", "tools": ["search", "get_fields"] } ],
  "agent_timeout": 450
}
JSON