Give the teammate you already work with a body.
Speak to your Reachy Mini and your Fleece AI teammate answers out loud — the same brain, memory and persona it has on WhatsApp, Slack, email or the phone. And the same tools: ask it to check the calendar or draft the email, and it goes and does it.
You
Did anything come in about the Tuesday deadline?
Your teammate
Two things did — I'll put them in your inbox now so you have them before the standup.
Ran its own tools to answer Nodded while it spoke Remembers the standup you mentioned yesterday
Same brain, one more channel
Your Reachy Mini becomes a channel of your Fleece AI Team workspace, alongside the seven it already answers on. Nothing to rebuild and nothing to keep in sync — ask it about something you discussed over chat this morning and it remembers.
It does not only talk
Link it to a Fleece AI agent and the robot runs the real thing: its tools, skills and integrations. It carries the task out rather than describing how you might.
Gestures come from the agent
Every reply carries an emotion, played as a recorded move from Pollen's library — a nod, a shrug, a small dance — while the robot speaks, as people do.
Four steps, one of them required.
-
Install the app
Find Fleece AI Teammate in the app store of your Reachy Mini dashboard.
-
Connect a Robot channel
In Fleece AI Team, open your agent, add a Robot channel, and copy the webhook URL it gives you. Paste your own Fleece AI API key there too and the robot runs the agent's tools; leave it empty and it still holds a conversation.
-
Paste it into the settings
Open the app's settings from the dashboard and paste that URL under Setup. That is the one required setting — you can already talk to your teammate from the page.
-
Give it a voice
To speak rather than type, add a Mistral key for hearing and an ElevenLabs or OpenAI key for the voice. With ElevenLabs it keeps its own cloned voice — the one it uses on calls.
Your keys stay on the robot
Settings live in a file only the robot's owner can read, and secrets are never sent back to the browser. Audio goes to the speech services you configured and to nothing else. Conversation history lives in your own workspace. Without voice keys the app still works — type from the settings page, or run it against the simulator, no hardware required.