/api/v1/agent/runExecute an agent template against a user message. The response includes the model output, an audit ID for the immutable audit log entry, DLP policy flags, and token usage. For healthcare/behavioral-health verticals a signed BAA must be on file; input containing PHI is blocked at the DLP layer and never forwarded to the model.
Request body
{
"agentId": "helpdesk-tier1-triage",
"message": "Router at floor 3 is offline since 09:00 — ticket ITSD-4421",
"systemPromptOverride": "(optional) Custom system prompt for advanced integrations."
}
// agentId — string (1–100 chars). Built-in templates:
// "coleman-clinical-docs" — Clinical documentation (Healthcare)
// "helpdesk-tier1-triage" — IT helpdesk triage + priority classification
// "compliance-evidence-collector" — SOC 2 / HIPAA / ISO 27001 evidence gathering
// "audit-trail-narrator" — Plain-language audit-log summaries
// message — string (1–32 000 chars). Never stored; only SHA-256 is written to the audit log.
// systemPromptOverride — string (max 8 000 chars). Overrides the template default prompt.Response
// HTTP 200
{
"response": "P2 — Network / Infrastructure. Recommended path: on-site check of patch panel ...",
"audit_id": "aud_01JQABCDE12345XYZ",
"policy_flags": [],
"model_used": "claude-haiku-3-5",
"tokens_used": 312
}
// HTTP 422 — PHI detected
{
"error": "Input blocked: protected health information (PHI) detected. Please remove PHI before submitting.",
"code": "DLP_INPUT_BLOCKED",
"policy_flags": [
{ "type": "PERSON_NAME", "offset": 12, "length": 9 },
{ "type": "DATE_OF_BIRTH", "offset": 34, "length": 10 }
]
}