joinly

Why we included this project

joinly is middleware that lets an AI agent sit in on a live video call and do real work, not just watch. It runs as an MCP server, connects to Google Meet, Zoom, and Microsoft Teams, and gives the agent meeting tools and resources to answer questions, take notes, or trigger actions in other systems while the call is happening. The setup is modular: you bring your own LLM, including a local Ollama instance, choose your speech-to-text and text-to-speech providers, and can attach extra MCP servers for things like web search or Notion editing mid-call. Teams that want a self-hosted, vendor-neutral way to put an agent on their calls will find the Docker-based install straightforward, and the built-in conversational flow handles interruptions and multi-speaker talk so the agent doesn't talk over people.