zclaw: personal AI assistant in under 888 KB, running on an ESP32
github.com
February 21, 2026
3 min read
🔥🔥🔥🔥🔥
61/100
Summary
zclaw is a personal AI assistant designed for ESP32 boards, with a firmware size limit of 888 KiB. It supports GPIO control, scheduled tasks, persistent memory, and allows for custom tool composition using natural language.
Key Takeaways
- zclaw is an AI personal assistant designed to run on ESP32 boards with a firmware size limit of 888 KiB, including all necessary components.
- The system supports features such as GPIO control, scheduled tasks, persistent memory, and integration with AI providers like Anthropic and OpenAI.
- Users can interact with zclaw via Telegram or a hosted web relay, and it includes built-in tools as well as user-defined tools for customization.
- The recommended starter board for zclaw is the Seeed XIAO ESP32-C3, and it has been tested on ESP32-C3, ESP32-S3, and ESP32-C6 variants.
Community Sentiment
Positives
- The potential for a personal AI assistant on an ESP32 opens up exciting possibilities for low-cost, accessible AI applications in everyday devices.
- The idea of creating an intelligent version of a Tamagotchi using minimal hardware showcases innovative uses of AI in personal projects.
Concerns
- There is disappointment regarding the reliance on LLM backends, as many users were hoping for local inference capabilities to enhance privacy and control.
- The lack of options for minimal assistants that work with locally hosted models highlights a significant gap in the current AI landscape.