Skip to main content

Agents are where business policy meets voice execution.

An agent in Callaro bundles the voice, script logic, objection handling, retrieval permissions, tool access, and response style that should govern a conversation.

Human answer

The agent moves into discovery, qualification, objection handling, and next-step capture.

Voicemail

The agent follows a different branch, keeping the message concise and aligned with campaign policy.

Conversation modes

The agent Advanced step can expose a conversation mode when your role allows it:
  • Pipeline — STT → LLM → TTS. This is the default production path for most campaigns.
  • Realtime — a speech-to-speech path used when the tenant/runtime is configured for live audio models.
Call logs can filter by this mode. Treat mode as an operational setting, not a separate API product: campaign launch, webhooks, and voice_sessions stay the same.

Prompt and policy guardrails

  • Keep one primary objective per agent (for example qualify + book).
  • Define explicit objection boundaries and escalation behavior.
  • Restrict tool access to only required CRM/calendar actions.
  • Review transcript samples weekly and tune branch logic iteratively.

Anti-patterns to avoid

The highest-performing agents usually have narrower objectives than teams expect at first. A script designed to qualify and book a meeting will outperform one that also tries to educate, negotiate, and close in a single call.