> ## Documentation Index
> Fetch the complete documentation index at: https://docs.medera.info/llms.txt
> Use this file to discover all available pages before exploring further.

# Agentic Framework FAQ

> Frequently asked questions

**Can I use my own LLM?**
Enterprise tenants can plug a customer-hosted LLM under BAA. Standard tenants use the Medera-managed inference layer (OpenAI `Medera reasoning` for voice agents, `gpt-4o` for SOAP / analysis, `gpt-4o-mini` for clinical NLP). **Can I write a custom agent?**
Yes — use the [Agent Builder Canvas](/agentic/canvas/overview) to compose from existing node types, or extend the Care Orchestrator with a new `agent type` and a corresponding `NodeRunner`. **Are agents deterministic?**
No — agents are LLM-driven. Every output is reported with calibrated confidence; every claim ties back to a span or citation. **What happens on crisis-language detection?**
The guardrails layer detects crisis language and immediately invokes the Crisis Response Agent with no gating. The C-SSRS Risk Expert is consulted for risk band classification. **Where can I see audit logs?**
PHI access logs: `GET /api/audit/access-logs`. Deployment actions: `deployment_audit_log` (WORM, exposed via [Audit API](/api-reference/audit/access-logs)).