How to Self-Host Vexa with Docker: Real-Time Meeting Transcription

This guide walks you through deploying Vexa yourself with Docker, then streaming your first real-time transcript. The whole stack is Apache-2.0 licensed, so everything you see here stays under your control.

Step 1 — Pull the Image

Vexa publishes an official lightweight image on Docker Hub. Start by pulling it and creating a data volume:

docker pull vexaai/vexa-lite:latest
docker volume create vexa-data

Step 2 — Run the Container

docker run -d --name vexa -p 8080:8080   -v vexa-data:/data vexaai/vexa-lite:latest

The API listens on port 8080. Mounting /data keeps transcripts and configuration persistent across restarts.

Step 3 — Configure Auto-Join Bots

Create a bot for each platform and hand Vexa a meeting link. The bot joins automatically, records the session, and exposes the transcript through a WebSocket endpoint your application subscribes to in real time.

Step 4 — Connect AI Agents via MCP

Vexa ships an MCP server, so agents built on the Model Context Protocol can search and summarize transcripts without custom glue code. Point your MCP client at Vexa and meeting knowledge becomes a first-class tool for your agents.

Wrap-Up

In four steps you have a self-hosted, real-time meeting transcription API covering Meet, Teams and Zoom — with MCP built in. For the full requirements and ready-to-use compose file, check the tool page below.