Skip to main content
This guide is for product and clinical teams new to LLM agents.

​
What is an agent?

In Medera, an agent is a task-specific actor with:
  1. A typed configuration — practice template, screener set, escalation policy, voice settings
  2. A documented execution flow — every step is in this documentation
  3. A bundle of typed experts — PHQ-9, C-SSRS, ICD-10, PubMed, Payer Policy, …
  4. Compliance guarantees — e.g. crisis-language detection bypasses agent gating

​
Why this matters

Generic LLMs hallucinate clinical claims. Medera agents:
  • Anchor claims to quoted spans, screener scores, or peer-reviewed citations
  • Report confidence in calibrated bands (VERY_HIGH → VERY_LOW)
  • Surface unavailable rather than fabricated values when signal quality is low
  • Log every step to an immutable audit trail (deployment_audit_log for deployment actions, phi_audit_logs for PHI)

​
First steps

  1. Pick an agent from the Agent Library
  2. Read the agent’s configuration section
  3. Use the Canvas to compose a workflow that includes the agent
  4. Publish a version, deploy to sandbox, and execute via /agentic/quickstart