Vexa-ai/vexa

What it solves

Vexa provides a self-hosted, open-source alternative to cloud-based meeting AI tools. It solves the problem of data privacy and vendor lock-in by allowing teams to capture meeting transcripts and build a persistent, owner-controlled knowledge base without sending conversations to a third-party cloud.

How it works

Vexa uses a bot fleet that joins Google Meet, Microsoft Teams, and Zoom calls to stream speaker-attributed transcripts in real time. These transcripts are compiled into a "workspace"—a git repository of Markdown files (following the Open Knowledge Format).

To process this knowledge, Vexa employs a sandboxed agentic runtime. This runtime spawns ephemeral, isolated containers (via Docker, Kubernetes, or local processes) to run AI agents that can read and write to the workspace, allowing them to compile raw meeting data into structured entity pages (people, projects, decisions) rather than relying on simple RAG over raw documents.

Who it’s for

• Regulated Enterprises: Organizations in banking, healthcare, or government that require air-gapped, sovereign AI intelligence.
• Developers and Teams: Those who want to own their meeting data as "knowledge as code" (portable Markdown files) and use AI agents to maintain it.
• API Users: Developers who only need a real-time transcription API for meeting bots without the agent layer.

Highlights

• Multi-platform Support: Joins Google Meet, Microsoft Teams, and Zoom (Jitsi support is in validation).
• Sovereign Data: Fully self-hostable and air-gap ready, ensuring data never leaves the user's perimeter.
• Knowledge as Code: Meetings are stored as diffable, greppable Markdown files in a git repo.
• Sandboxed Execution: Agents run in isolated, ephemeral containers to ensure security when handling business data.
• Flexible Deployment: Supports single-container "Lite" installs, Docker Compose, or scalable Kubernetes (Helm) deployments.