Voice Agent with Render Workflows

Why deploy voice agent workflow on Render?

A voice agent workflow is an orchestration pattern that connects voice AI conversations to backend processing tasks. It enables real-time data collection through voice interfaces while triggering automated workflows (like claim verification, fraud checks, and notifications) based on conversation outcomes.

This template wires together four services—a React frontend, FastAPI backend, LiveKit voice agent, and Render Workflows orchestrator—with all inter-service communication and environment variables pre-configured. Instead of manually setting up LiveKit agent dispatch, workflow task definitions, and real-time progress streaming between services, you get a working voice AI demo with one click. Render Workflows handles the background claim processing with parallel task execution and automatic retries, which would require significant infrastructure work to build yourself.

Architecture

What you can build

After deploying, you'll have a working voice AI demo where users can call a browser-based agent to file an insurance claim by describing their situation. The agent collects the relevant details, then hands off to background workflow tasks that run claim processing steps—policy verification, damage analysis, fraud checks, cost estimates, and repair shop lookup—with live progress visible in the UI.

Key features

• LiveKit Voice Integration: Browser-based voice calls connect to a LiveKit agent using OpenAI GPT-4o for conversation, Whisper for speech-to-text, and OpenAI TTS for responses.
• Render Workflows Orchestration: Background claim processing uses @app.task decorators with asyncio.gather to run independent steps like damage analysis and fraud checks in parallel.
• Real-time Progress Updates: The React frontend displays live task progress as workflow subtasks complete, showing claim status through each processing stage.
• Blueprint Infrastructure-as-Code: A render.yaml file defines all four services (frontend, API, agent, workflows) for one-click deployment with environment group references.
• Docker Compose Dev Setup: Local development runs API, agent, and frontend via Docker Compose while the workflow dev server runs separately with the Render CLI.

Use cases

• Insurance startup prototypes automated claims intake with voice AI
• Developer demos real-time workflow orchestration for customer service calls
• Product team tests parallel background task processing during live conversations
• Fintech engineer builds voice-driven support with fraud detection pipeline

What's included

Next steps

1. Open the frontend URL and start a voice call with the AI agent — You should hear the agent greet you and ask for your phone number to begin the claim process
2. Test a complete claim by providing sample info (phone number, location, damage description, zip code) and ending the call — You should see the claim progress UI update in real time as each workflow task runs: policy verification, damage analysis, fraud check, estimate generation, and repair shop recommendations
3. Configure the three environment groups (livekit-config, render-config, ai-config) in the Render Dashboard under Env Groups — After adding the variables and redeploying, the voice agent should connect successfully and OpenAI-powered speech recognition and responses should work

Resources

• Render Workflows official site