StandIn
Your AI teammate in Microsoft Teams calls
12 followers
Your AI teammate in Microsoft Teams calls
12 followers
StandIn puts your own AI agent into real Microsoft Teams calls: it answers, talks back in real time, watches shared screens, and shows up as an animated avatar tile. Try it in seconds in the free sandbox (no Azure, no setup), or connect your own bot on the free developer tier. The Teams Voice plugin is open source and runs your AI on your own infrastructure, so your keys, prompts, and transcripts never leave your machine.
Interactive
Free Options
Launch Team / Built With
The Teams Voice plugin being open source is a really nice touch for privacy. One thing that would make the sandbox way more useful is letting me upload a short voice sample so the avatar matches my actual tone and pacing, instead of the default voice. Would help me show the team what real deployment could feel like before I commit to setting up the self-hosted version.
@mayimbialik Thanks, glad the open source part landed.
On the voice: you can already do this today, it just happens outside the plugin. Prepare your voice with ElevenLabs, Deepgram, or Cartesia and point the plugin at it, so the avatar runs with your tone and pacing instead of a default.
The part worth showing your team is the media layer. Most open source Teams integrations are text chat only. This one enables full voice and video calls inside Teams: CVI, the agent seeing you or your screen, real time back and forth. You build the brain with OpenClaw, Hermes, or LiveKit, and the plugin handles the media connectivity into Teams.
That should get the sandbox demo much closer to what real deployment feels like.
@mayimbialik Adding to this, since the voice question is really a question about what the demo can show. Once the agent is an actual participant on the call instead of a chat bot, these open up:
Meeting copilot: it joins as a video participant, watches the screen share, and knows who is presenting what. Ask it "what did the earlier slide say" and it answers from keyframe history rather than only the current frame.
Live troubleshooting: share your screen, it reads the error and talks you through the fix out loud.
Watch it work: give it a task mid-call and it shares its own screen back to you, streaming screenshots and generated images onto its video tile with captions while it runs. You see the work happen instead of waiting for a text summary.
Call me back when done: hand it something long, keep the meeting moving, and it phones you with the result when it finishes. Or tell it to just update you, and the answer comes back however you asked for it, chat message or voice call. Unanswered calls land as a Teams voicemail with transcription.
Chat to call bridge: ask in a Teams chat, get the answer back as a phone call.
Customer-facing CVI: an agent people can video call, with a lip-synced avatar, expressions, and gaze toward whoever is speaking.
The brain stays yours, OpenClaw, Hermes, LiveKit, whatever you already run. The plugin is the media layer that carries it into a Teams call with voice, video, and screen awareness.
honestly the sandbox is wild, i dropped it into a test call and it actually picked up the screen share and responded without me touching anything. avatar tile is a nice touch too
@tansu750955 Thanks, that is exactly the moment we built it for. No prompting, no setup, it just notices the screen share and joins in.
If you want the full picture, try the community and developer package. It is free and installs into your own Teams tenant. The main difference from the sandbox is that it connects calling and messaging together rather than calling alone. Once both paths are live, the connected agent, whether that is OpenClaw, Hermes, ElevenLabs, or LiveKit, processes every incoming input instead of just what happens on one call.
That is what unlocks the cross-channel behavior: it can message you the meeting minutes afterward, take on a task during a group meeting and report back when it is done, or share files into the chat to support what is being discussed live.
the sandbox with no Azure setup is a really smart move, gets you into a real Teams call in seconds and that's when the avatar tile concept actually clicks. nice execution.
@thfomas_cooper Thanks, that was the whole bet. Nobody quite believes the avatar tile until they are sitting in a call with it, and asking people to do tenant configuration first meant most never got that far.
Since you got there quickly, a question worth your take: we are weighing a managed app bot listed on the Teams store, working the way the Jira bot does, where you add it yourself and it just runs. No Azure setup, no infrastructure, nothing to stand up. Would that be the version you would reach for, or would you rather keep it self-hosted so the data path stays inside your own tenant?