# Egma

> Egma is a self-hosted, open-source platform for simulation testing and production monitoring of voice agents with LiveKit, Pipecat, and Retell support.

- **Category:** AI Chat
- **Licence:** Other
- **Latest version:** cli-v0.3.3
- **Open source:** oss
- **Languages:** Typescript, Python
- **Platforms:** Docker Compose
- **Tags:** Livekit, Openai, Docker Compose, Self Hosted, Pipecat, Voice Agents, Deepgram, Simulation Testing, Production Monitoring, Retell, Cartesia, AI Testing

## Description

Egma gives developers a way to test voice agents before release and watch them in production. You write simulation suites that run synthetic voice or text conversations against your agent, with mocked tool responses, then grade the results. The same platform ingests live traffic so you can review actual calls, tool invocations, and metrics. It works with LiveKit, Pipecat, and Retell frameworks. You supply your own model-provider keys — OpenAI, Cartesia, Deepgram — and pay those vendors directly. The stack runs in Docker Compose: PostgreSQL stores metadata, ClickHouse holds traces, MinIO keeps recordings, LiveKit handles media, Redis does signaling, and Egma's own API, web UI, simulator, and grader containers tie it together. A CLI drives the workflow. The code is MIT licensed except for ee/ folders.

## Links

- [Homepage](https://github.com/egma-ai/egma)
- [Website](https://app.egma.ai)
- [GitHub Repository](https://github.com/egma-ai/egma)
- [Discord Community](https://discord.gg/v4KHSNHfPj)
- [Egma Cloud](https://app.egma.ai)

---
Source: https://selfhost.directory/project/egma