Xiaomi Miloco

What Xiaomi Miloco does

Miloco uses video and audio from Mi Home cameras as a perception gateway and Xiaomi's in-house MiMo large model as its reasoning engine. Miloco 2.0, released on 2026-06-18, was re-architected as a plugin on OpenClaw. Its features include built-in common sense that detects hazards and raises tiered alerts, identity recognition that fuses face and body-posture signals, home memory of family members' habits and preferences, and household tasks that break vague, long-term goals into trackable steps.

Later releases added a rule and task re-architecture, Mi Home scene triggering from rule actions, experimental pet recognition, one-click self-upgrade from the dashboard, and token usage accounting by model and endpoint. The changelog credits release v2026.7.17 with a Hermes Agent compatibility layer for pluggable agent runtimes, along with dual-camera perception and a dedicated Tasks tab.

Key features

• Perception from Mi Home camera video and audio
• Hazard detection with tiered alerts from built-in common sense
• Identity recognition of family members from faces and body posture
• Home memory of habits and preferences used for proactive decisions
• Household tasks built from vague, long-term goals
• Hermes Agent compatibility layer for pluggable agent runtimes (v2026.7.17)

When to use it

• Raise an alert when a hazard such as a fall is detected on camera
• Personalise device control and reminders for each family member
• Turn a loose household goal into trackable tasks handled by the agent

Who it is for: Mi Home users and developers interested in a camera-driven, agent-based smart home.

How it fits with Hermes Agent

Runs on OpenClaw as a plugin, and release v2026.7.17 added a compatibility layer for Hermes Agent as a pluggable agent runtime.

Requirements: Mi Home cameras; OpenClaw, on which Miloco 2.0 runs as a plugin

FAQ