About the service
A workbench for support agents: one editor holds both agentic playbooks and deterministic step-by-step flows, wired up to a knowledge base, your own APIs, chat channels and telephony. Finished agents run on Voiceflow's own infrastructure, with transcripts, analytics and LLM-powered evaluations sitting next to the builder. Billing comes in two layers: the plan unlocks platform features and grants a monthly pool of credits, but the same credits are drained by orchestration — every user message, every minute of a call — and by model tokens: GPT from OpenAI, Claude from Anthropic, Gemini from Google, Llama and Qwen served through Groq, Voiceflow's own models, plus speech recognition and synthesis from ElevenLabs, Cartesia, Deepgram and others. When the pool runs out you buy credit bundles or turn on auto top-ups, so the monthly total depends on which model you dropped into the flow.