Give Your Agent
Eyes and Ears
Low-power perception, emotional interaction, and Agent connectivity — the "Stripe" for AI native devices.
The Problem
AI Agents Are Brilliant.
But They Can't See or Hear.
Vision & audio are prerequisites — not nice-to-haves — for Agents in the real world.
Agents can call APIs and run code — but they still can't see, hear, or speak. eyespot fills that gap.
The most critical AI infrastructure is open source. eyespot follows the same path.
Our Answer
Eyespot: an open-source hardware & software development platform
that lets Agents quickly access multimodal-interactive smart devices.
Real-World Scenarios
SPOT → Understand → Act
Someone at the door? Pet on the sofa? Package arrived? eyespot lets your Agent perceive all of this in real time — and act on it.
"Anyone at the door?"
👁 Perceives: Doorbell camera detects a delivery person placing a package.
Agent → Sends notification + Unlocks smart lock
"How's grandma doing today?"
👁 Perceives: Camera detects prolonged inactivity — no movement for 2 hours.
Agent → Calls family + Returns AI-generated daily timeline
"Always watching. No prompt needed."
👁 Perceives: Camera sees you grab your keys at 8:02 AM.
Agent → Triggers 'leaving home' mode — lights off, windows closed
"Watch over the kids."
👁 Perceives: OvO camera reads facial expression — child is crying.
Agent → Switches to comfort mode + Notifies parent
"What does she want to do now?"
👁 Perceives: Robot detects high energy + laughter in the last 5 min.
Agent → Suggests a game, not a story — no manual switch needed
"Keep the pet company."
👁 Perceives: Microphone detects dog whining — alone for 3 hours.
Agent → Initiates play interaction + Logs behavior to owner
"What does this menu say?"
👁 Perceives: AI glasses parse foreign-language menu in first-person view.
Agent → Reads translated items aloud in real time
"Summarize what they said."
👁 Perceives: Pendant microphone transcribes 10-min conversation continuously.
Agent → Returns 3-bullet summary on demand
"Coach my pronunciation."
👁 Perceives: Device hears mispronounced word + reads visual scene context.
Agent → Corrects in real time + Generates situational drill
MULTIMODAL INTERACTION INFRASTRUCTURE
Hardware as Native Skills, for 🦞OpenClaw
THREE-LAYER ARCHITECTURE
The Eye
Perception Layer
Camera/ToF + mic array + speaker + night vision + edge AI chip
The Brain
Understanding Layer
Local multimodal VLM (0.5T–6T) + scene-specific lightweight models
The Bridge
Connection Layer
Agent framework adapter + RESTful API + MCP protocol + WebRTC + MQTT
Hardware as Native Skills
eyespot packages cameras, microphones, and speakers as first-class Agent Skills. Three layers: The Eye (camera/ToF + mic + speaker), The Brain (local VLM, 0.5T–6T compute), The Bridge (Agent framework adapter). Your Agent calls them the same way it calls any tool — no custom drivers, no glue code.
Plug In Any Hardware
ONVIF cameras, RTSP streams, WebRTC endpoints, USB microphones, Bluetooth speakers — eyespot abstracts them all behind a unified API your Agent already understands.
Flexible Inference: Local or Cloud
Local inference on capable hardware (0.5T–6T) — data stays on device, zero cloud dependency. Cloud inference for cost-sensitive deployments. Choose based on your hardware spec and privacy needs.
BUILT-IN SKILLS
Full-stack perception, understanding, and action — ready to use out of the box.
Scene Observation
Ask in natural language, get real-time scene descriptions.
Object Recognition
Identify people, pets, packages — structured JSON output.
Event Detection
Always-on monitoring, proactive event push to Agent.
Visual Tasks
Set complex tasks with visual conditions via natural language.
Timeline Recall
AI-understood event timeline, not traditional video playback.
Open API
RESTful API, Webhook, deep Home Assistant integration.
Application Scenarios
Three Hardware Categories
Through the Eyespot framework, three major hardware categories achieve a complete physical-world perception loop: SPOT → UNDERSTAND → ACT.
Smart Home Terminal
SMART HOME TERMINAL
AI Security Camera
On-device NPU for real-time fall detection & intrusion recognition, local inference with zero cloud upload
AI Smart Doorbell
Face recognition + voice intercom, automatic stranger alerts, delivery auto-logging
Elderly/Child Care Camera
Proactive anomaly sensing, inactivity alerts, remote family real-time linkage
AI Companion Robot
AI COMPANION ROBOT
OvO Desktop Robot
In-house reference hardware with camera + screen + voice all-in-one, a family AI assistant that gets things done
AI Toy / Children's Companion
Visual emotion sensing, adaptive dialogue content, engaging smart interactive experience
Pet Companion / Monitor Robot
Pet behavior recognition + remote interactive feeding, real-time companionship when you're away
Smart Wearables
SMART WEARABLES
AI Smart Glasses
First-person visual understanding, real-time scene parsing, seamless natural voice interaction
AI Pendant / Wearable Agent
Ultra-compact edge AI terminal, a portable Agent perception node that captures life's key moments
AI Language Practice Device
Real-time speech recognition + visual context sensing, immersive language learning & oral practice
Covering 0.5T–6T full compute range, with OvO reference hardware for rapid development and mass production.
HARDWARE INQUIRY
AI Developer or Hardware Manufacturer? Let's Talk.
Whether you're interested in OvO hardware procurement, ODM customization, or enterprise-scale deployment — we'd love to help you find the right solution.
ECOSYSTEM
Works With Any Agent Framework
eyespot is framework-agnostic. If your Agent can call tools, it can use eyespot Skills.
TECHNOLOGY COMPARISON
Eyespot vs XiaoZhi AI vs LiveKit
Three projects with fundamentally different positioning. XiaoZhi AI is a voice companion, LiveKit is A/V transport infrastructure, and eyespot is the only open-source framework that packages hardware perception as Agent Skills.
| Dimension | XiaoZhi AI | LiveKit | eyespot |
|---|---|---|---|
| Positioning | Voice Companion | Real-time A/V Cloud Infra | Hardware Perception Skill Framework |
| Visual Perception | |||
| Audio Perception | |||
| Structured JSON Output | |||
| Proactive Event Push | |||
| Hardware Product | ESP32 ($7 DIY) | None | OvO Series |
| Agent Framework Integration | Weak | Voice Agent Pipeline | Native Skill Calls |
| Open Source License | MIT | Apache 2.0 | MIT |
| Business Model | Ecosystem + HW Licensing | Open Core + Usage Cloud | Open SDK + HW + Cloud |
| Valuation / Funding | Shifang Ronghai Ecosystem | $1B (Series C) | Pre-Seed |
Key Differentiator
XiaoZhi AI gives Agents a voice, LiveKit gives Agents A/V transport, eyespot gives Agents the ability to see, hear, and understand the physical world — complementary, not competitive. eyespot fills the perception layer gap.
BUSINESS MODEL
Open Source Free. Commercially Sustainable.
Core SDK is free and open source forever. Monetization through hardware sales, pay-per-analysis Agent Deployment, and enterprise plans. Contact us for specific pricing.
Open Source SDK
Free ForeverMIT licensed. The core framework is fully open source. Any developer can use, modify, and distribute freely.
- eyespot Core SDK
- MCP Server
- Community Support
- GitHub Docs
OvO Hardware
One-Time PurchaseOfficial reference hardware built on eyespot — a plug-and-play flagship Agent Skill.
- Lite / Pro / Ultra tiers
- Pre-installed eyespot
- Ready to ship
- Hardware warranty
Agent Deployment / Managed Model
Pay Per AnalysisManaged inference, device management, and event stream processing. No infrastructure to build — pay only for what you use.
- Managed inference
- Fleet management
- Event stream processing
- SLA guarantee
Enterprise
Custom PlanPrivate deployment, custom integration, and dedicated support for large-scale deployments and compliance needs.
- Private deployment
- Custom integration
- Dedicated support
- Compliance certification
GET PRICING
Want specific pricing? Let's talk.
Whether it's Agent Deployment pay-per-analysis plans, hardware bulk procurement, or enterprise private deployment — we'll tailor a solution for you.