Vexa-ai / vexa

Self-hosted meeting bot that transcribes calls in real time and feeds sandboxed AI agents — your meetings, your data, your infrastructure.

活跃维护 Apache 2.0 Python Tracked
2.7k 448 4 小时前
CINPM
google-meet meeting-assistant meeting-minutes meeting-notes ms-teams notetaker zoom ms-teams-app api mcp self-hosted speech-to-text transcription

星标趋势

数据积累中,暂无足够数据生成趋势图

AI 分析

项目摘要

Vexa is an open-source, self-hosted meeting bot and transcription API that joins Google Meet, Microsoft Teams, and Zoom calls to stream speaker-attributed transcripts in real time. It compiles meetings into a Markdown knowledge base and includes a sandboxed agent runtime with MCP server support for AI agents to process meeting content.

为什么值得关注

Vexa inverts the SaaS meeting-AI model by letting teams self-host the entire stack, own their data, and integrate with their own models — while still offering a hosted option. The combination of real-time bot-based transcription, knowledge-as-code Markdown output, and a sandboxed agent runtime with MCP support makes it a unique full-stack solution for meeting intelligence.

优势

  • Self-hosted, air-gap-ready architecture with full data ownership
  • Real-time speaker-attributed transcription across Google Meet, Teams, and Zoom
  • Integrated sandboxed agent runtime with MCP server for AI agent workflows
  • Knowledge-as-code approach with Markdown output in git repos
  • Active development with 20 releases in 6 months and 32 contributors

局限性

  • No automated tests despite CI being present
  • No Docker support or examples for easier onboarding
  • Relatively new project (created Feb 2025) with limited production track record
  • Self-hosting requires GPU for transcription or dependency on vexa.ai tokens

使用场景

  • Self-hosted meeting transcription and minute generation for privacy-conscious teams
  • Building AI agent workflows on top of meeting data via MCP server
  • Creating a searchable, version-controlled knowledge base from meeting conversations
  • Regulated enterprise environments requiring air-gap deployment
目标用户: Engineering teams and organizations that want self-hosted meeting intelligence with data sovereignty, and developers building AI agent workflows on meeting data
学习曲线:
分析模型:LongCat-2.0 | 分析时间:1 个月前