How Paladin Works

Paladin is a platform for building and running voice AI agents. You define a conversation flow, connect a phone number, and Paladin handles the rest — transcribing the caller's speech (STT), generating intelligent responses (LLM), speaking them back in a natural voice (TTS), and returning structured results when the call ends.

The core loop#

sequenceDiagram
    participant U as Dashboard / Your App
    participant API as Paladin API
    participant STT as Transcriber (STT)
    participant LLM as LLM Provider
    participant TTS as Voice Synthesizer (TTS)
    participant Tel as Telephony Provider
    participant Cal as Caller / Contact

    U->>API: Trigger call (dashboard or API)
    API->>Tel: Initiate outbound call
    Tel->>Cal: Phone rings
    Cal-->>Tel: Answers
    Tel-->>API: Raw audio stream
    loop Conversation
        API->>STT: Caller audio
        STT-->>API: Transcribed text
        API->>LLM: Transcript + agent prompt + context
        LLM-->>API: Agent response text
        API->>TTS: Response text
        TTS-->>API: Synthesized audio
        API->>Tel: Audio stream
        Tel->>Cal: Agent speaks
    end
    API->>API: Extract context, run webhooks
    API-->>U: Run record (transcript, recording, gathered data)

Key components#

Workflows (Agents) The conversation logic. A workflow is a graph of nodes (conversation steps) connected by edges (conditional transitions). You define what the agent says, when it moves on, and what data it collects.

Runs Every execution of a workflow creates a run. The run record holds the transcript, recording, extracted data, and cost information.

Telephony The phone infrastructure. Paladin connects to your telephony provider (Twilio, Vonage, etc.) to place and receive calls. The audio streams between the caller and Paladin in real time.

Transcriber (STT) Converts the caller's speech to text in real time. Paladin sends the audio stream to your configured speech-to-text provider and uses the transcript to drive both the LLM and the final run record.

LLM Provider Processes the transcript and the active node's prompt to generate the agent's next response. It also evaluates edge conditions to decide when to move the conversation forward.

Voice Synthesizer (TTS) Converts the LLM's text response to audio and streams it back to the caller. The choice of TTS provider and voice is configurable per agent.

How it fits together#

When you trigger a call:

1. Paladin instructs your telephony provider to dial the number
2. When the caller answers, a real-time audio pipeline opens
3. The caller's speech is transcribed by the STT provider
4. The transcript is sent to the LLM with the active node's prompt and conversation history
5. The LLM responds — the response is synthesized to audio by the TTS provider and streamed to the caller
6. When an edge condition is met, Paladin transitions to the next node
7. When an end node is reached, the call ends
8. Post-call: context is extracted, webhooks fire, the run record is saved

Next steps#