Vexa-ai/vexa meeting transcription API

I came across Vexa-ai/vexa on GitHub Trending: an open-source meeting transcription API that supports Google Meet, Microsoft Teams, and Zoom, with auto-join bots and real-time transcripts streamed over WebSockets. It also includes an MCP server so you can wire the transcript stream into AI agent workflows without building a bunch of glue yourself.

This is the kind of tool I’d reach for when I need reliable “notes → searchable context → actions” from live meetings, especially if the meetings happen inside corporate platforms where direct audio capture isn’t practical. One concrete setup: spin up your own self-hosted transcription service, have an auto-join bot join scheduled calls, stream partial transcripts to your backend, and feed completed segments into a RAG index or an agent that extracts decisions, tasks, and action items.

What to look at first:

- The repo entry point (README) to confirm the exact deployment mode you want: self-host vs hosted SaaS - The WebSocket transcript interface docs, since that’s the core integration surface - The MCP server section, which is the quickest way to connect into agent tooling - Any configuration/credentials notes for supported meeting providers (Meet/Teams/Zoom)

Overall, it’s a production-shaped piece of infrastructure: real-time streaming, API-first access, and an agent integration path via MCP. If you’re already building agent systems, this should plug into your pipeline faster than rolling your own transcription + platform connectors.