edge-ai-agents

MimiClaw: Harness on a $5 chip. No OS(Linux). No Node.js. No Mac mini. No Raspberry Pi. No VPS. Hardware agents OS.

Run a <400ms latency Voice Agent on just 4GB VRAM. Fully offline, no API keys required. Optimized for GTX 1650 and edge robotics with zero-copy inference. (Apache 2.0)

On-device AI recipes for building private, real-time applications with production-grade voice, language, and vision SDKs. Ready-to-use, open-source code examples.

A high performance C++ SDK for AI Agents

Local AI Agent firmware running on ESP32-S3, integrating offline voice wake-up with cloud TTS, supporting local LLM inference, tool calling, long-term memory storage and autonomous task execution.

Curated list of papers, frameworks, benchmarks, and applications for multimodal AI agents (LLMs, text-to-image, speech, world models, etc.) on mobile and edge devices.

InferNode is a security-focused 64-bit Inferno® OS (ARM64/AMD64) for embedded systems, servers, and AI agents. GPL-free, headless-capable, with 280+ utilities and 9P filesystem protocol. Providing a namespace-based alternative to MCP servers. Namespace-bounded containment of AI agents.

Python SDK for WildEdge

[~] IntegratedPipeline is a standalone AI Agent Library for Non-LLM memory Augmented Agentic Framework, Specifically designed to provide Agentic capability for any Autonomous Agentic Framework locally and Coordinatively that runs efficiently on small embedded systems to High end embedded systems, While continuously learning and runs efficiently.

Sub-3B local models drive a real desktop browser from three Android phones (2017, 2021, 2022, two SoC vendors) via structured page perception. 15 models, verified tasks, raw-HTML control, full JSONL logs, offline replay.

Swift SDK for WildEdge

Android SDK for WildEdge

NeuroStride is an AI-powered neurorehabilitation platform combining BCI-based robotic control, real-time exercise coaching, clinical AI assistance, multilingual voice support, and smart pharmacy integration to make neurological recovery accessible from home.

Node runtime for Harness Development Kit— bindings for liblloyal C++ kernel and llama.cpp

《拓扑生成范式：算力革命与黑箱破解》针对现有AI拟合路线算力爆炸、黑箱不可解释、沙箱越狱问题，提出低维骨架+表层填充的范式迁移，可大幅降低算力，实现生成过程可追溯可控。Topology‑Generating Paradigm: skeleton‑fill two‑layer forward‑generation framework. Solve AI black‑box, jailbreak and computing‑power explosion problems. Reduce computational cost by orders of magnitude, interpretable & traceable generation.

Edge-based fall detection system on Raspberry Pi 5 — YOLOv8 + MoveNet + real-time web monitor + WeChat alerts

⚙️ Build and train flexible generative recommendation models with modular tools, multi-GPU support, and integration with Hugging Face and LlamaFactory.

My homemade autonomous AI robot: What happens when you put a pan-tilt camera, speaker & mic, tof distance sensors, environmental sensors, and more on omni-directional wheels and then let an AI control the body?

Resource-Aware Agentic Swarm

Edge AI platform on RK3588S using Linux + Zephyr AMP, with local AI agents, inter-core IPC, and real-time GPIO/I2C/PWM control.