Recipes
Updated
Agora Recipes is a growing catalog of runnable examples, integration patterns, and end-to-end references across Agora products.
Quickstart
8Voice Agent quickstart
Start with the shortest path to your first Voice Agent.
Android Voice Agent
Build a native Android voice-agent client that owns RTC/RTM lifecycle and renders live transcripts.
Flutter Voice Agent
Run a cross-platform Flutter voice-agent app with RTC audio, RTM events, and transcript UI.
Go Voice Agent
Run a Go Gin backend with a Next.js voice-agent client and managed model defaults.
Next.js Voice Agent
Run a browser voice-agent starter with RTC audio, RTM events, server routes, and transcript UI.
Python Voice Agent
Run a Python FastAPI backend with a Next.js voice-agent client and managed model defaults.
React Native Voice Agent
Build a React Native voice-agent client that renders transcripts through Agora engine adapters.
SwiftUI Voice Agent
Build a native SwiftUI voice-agent client with RTC/RTM lifecycle and live transcripts.
Add features
14Content Filtering
Hear blocked sentences redacted by a custom LLM endpoint before they reach TTS.
Cross-Session Memory
Give the agent memory of prior conversations so it recognizes returning users.
Custom LLM
Swap the managed LLM for your own OpenAI-compatible endpoint while keeping managed STT and TTS.
Custom Modalities
Return audio from a custom endpoint so the agent skips the separate TTS step.
Dynamic Instructions
Change prompts, injected context, and runtime instructions without running a separate LLM service.
Dynamic Tool Sets
Run a smart-home agent whose available tools change by room and keyword scene.
Event Observability
Watch agent state, metrics, errors, and transcript events update in a live web timeline.
Filler Words & Graceful Exit
Add natural filler phrases during LLM latency and a graceful goodbye at session end.
Interruption Handling
Compare interruptible, locked, and keyword-triggered barge-in behavior for voice turns.
MCP Tools
Expose a FastMCP tool server that Agora invokes from a managed LLM conversation.
Retrieval-Augmented Generation
Ground replies in a small document corpus through a custom RAG endpoint.
Server-side Webhooks
Receive Agora server callbacks, verify them, store them, and stream them to a live timeline.
Speaker Lock
Focus the agent on the primary speaker and suppress background voices without enrollment.
Tool Calling
Run tools inside a custom LLM endpoint and persist note-taking results in SQLite.
Explore scenarios
8Agent Handoff
Try a travel concierge that moves between triage, booking, and support personas.
Real-time Translator
Translate spoken input into a target language and speak the translated response.
Real-time Vision (MLLM)
Use OpenAI Realtime with camera input so the agent can answer what it sees.
Real-time Voice (MLLM)
Replace the STT-LLM-TTS pipeline with an OpenAI Realtime voice-to-voice model.
Thymia Biomarkers
Explore a wellness demo that injects live voice biomarker scores into an AI therapist session.
Vision (Camera Input)
Publish camera frames to a managed vision-capable LLM and ask what it sees.
Voice RPG
Play a voice RPG where a Dungeon Master uses MCP tools for dice, combat, and inventory.
Wellness Coach
Run a wellness conversation stack with optional voice and video biomarker signals.