Voice AI Orchestration Framework

Build multi-role voice assistants without rebuilding the architecture.

Get Started

Start where we stopped.

Powered By

LiveKit for real-time voice, Zapier MCP for integrations, Mem0 for persistent memory — a multi-agent voice orchestrator without building the stack yourself.

LiveKit

Real-time voice and room orchestration

Zapier MCP

Calendar, booking, and external tools

Mem0

Conversation memory and context

Open Source

Built in the open.

Free to use. Community-driven. No platform lock-in.

Fork it. Improve it. Extend it.

View on GitHub

Get up and running.

From clone to running agent in 8 steps.

01

Get the repo

$git clone https://github.com/elhaweet/replaxy.git
$cd replaxy
02

Install dependencies

$uv sync
03

Initialize the project

$replaxy init
04

Configure credentials and integrations

$replaxy setup

Answer the prompts — LiveKit is required; Mem0 and Zapier MCP are optional.

05

Validate configuration

$replaxy validate

Optional but recommended.

06

Download runtime assets

$uv run python src/agents.py download-files

Only needed once.

07

Run the agent

$replaxy run

Use replaxy run --dev for local LiveKit testing.

08

Dispatch and connect

$uv run python scripts/dispatch_agent.py -r my-room

In another terminal. Then join the room (e.g. Agents Playground).

Optional

Customize agents

$cp config/agents.example.yaml config/agents.yaml

Edit config/agents.yaml (prompts, voices, handoffs) and run again with replaxy run.

Check integrations

$replaxy doctor
View on GitHubRead the Docs