PublishedGuidanceLAST REVIEWED · 30 AUG 2026
Intelligence Execute API
Send an explicit normalized task and control-plane constraints.
POST /intelligence/execute
{
"task": {
"input": "Assess this change request"
},
"objective": "balanced",
"risk": "medium",
"constraints": {
"max_cost_usd": 0.05,
"max_latency_ms": 10000,
"privacy": "standard",
"allowed_regions": [
"us"
]
},
"evaluation": {
"required": true
},
"allow_escalation": true,
"allow_multi_model": false,
"stream": false
}Execution boundary
Current strategies are single-model and confidence escalation. Do not treat allow_multi_model as proof that automatic composition is generally available.
Was this page helpful?