Developer tnm has released zclaw, a personal AI assistant that runs entirely on ESP32 microcontrollers with a firmware footprint under 888 KB. Written in C, the open-source project supports scheduled tasks, GPIO control, persistent memory, and natural language tool composition. The assistant integrates with popular AI providers including Anthropic, OpenAI, and OpenRouter through both Telegram and web relay interfaces.

Hardware compatibility spans ESP32-C3, ESP32-S3, and ESP32-C6 boards, with other ESP32 variants expected to work after manual ESP-IDF target setup. The project includes comprehensive tooling for building, flashing, provisioning credentials, and benchmarking latency. Users can install via a one-line bootstrap script or through manual installation, with options for secure credential storage in flash memory. The development team recommends the Seeed XIAO ESP32-C3 as a starter board.

The project has gained early traction with 201 stars on GitHub and includes features like timezone-aware scheduling, custom tool definitions, and GPIO read/write operations with built-in guardrails. Performance benchmarking tools measure both relay and direct serial communication paths. Complete documentation and guides are available through the project's docs site, covering everything from initial setup to advanced configuration.