GitHub - tnm/zclaw: Your personal AI assistant at all-in 888KiB (~25KB in app code). Running on an ESP32. GPIO, cron, memory, and more.
Hacker News
February 21, 2026
AI-Generated Deep Dive Summary
GitHub introduces zclaw, a lightweight AI assistant designed to run on ESP32 microcontrollers with an impressive all-in firmware size of 888 KiB, making it one of the smallest personal AI assistants available. Built in C, zclaw is optimized for ESP32 boards like the ESP32-C3 and ESP32-S3, offering features such as scheduled tasks, GPIO control, persistent memory storage, and natural language processing through a Telegram interface or web relay. Its compact size includes everything from the AI logic to Wi-Fi networking, TLS encryption, and certificate bundles, ensuring it stays within strict hardware constraints.
The project’s significance lies in its ability to bring AI functionality to low-power, embedded devices without sacrificing performance. zclaw supports multiple AI providers, including OpenAI, Anthropic, and OpenRouter, allowing users to integrate custom tools through natural language commands. Its GPIO control features guardrails for safety, enabling interaction with hardware components while maintaining stability. Additionally, its persistent memory capabilities ensure data survives reboots, making it ideal for IoT applications requiring reliability.
zclaw’s design caters to makers and hobbyists looking to experiment with AI on small devices, offering tools like cron-style scheduling for periodic tasks and a web-based relay interface for remote interaction. Its modular approach allows users to provision encrypted credentials securely and reset settings as needed, simplifying setup while maintaining flexibility. The project also includes detailed documentation and scripts for quick installation, testing, and benchmarking, making it accessible even to those new to embedded systems.
For tech enthusiasts, zclaw represents a breakthrough in AI accessibility
Verticals
techstartups
Originally published on Hacker News on 2/21/2026