This project addresses the challenge of running advanced AI functionalities on highly resource-constrained embedded systems. It provides a personal AI assistant, zclaw, specifically designed for ESP32 microcontrollers, offering a suite of features within a remarkably small firmware footprint. The target audience includes embedded systems engineers, IoT developers, and hobbyists seeking to integrate intelligent, interactive capabilities into low-power devices, benefiting from a powerful assistant that requires minimal hardware resources.

How It Works

zclaw is implemented in C, adhering to a strict <= 888 KiB all-in firmware size target. This constraint encompasses not only the zclaw logic but also the ESP-IDF/FreeRTOS runtime, Wi-Fi/networking stack, TLS/crypto libraries, and certificate bundles. Its core approach leverages efficient C programming and careful resource management to deliver AI capabilities on microcontrollers. A key design element is its ability to compose custom tools through natural language commands, allowing users to extend its functionality dynamically.

Quick Start & Requirements

• Primary install: bash <(curl -fsSL https://raw.githubusercontent.com/tnm/zclaw/main/scripts/bootstrap.sh) or ./install.sh if cloned.
• Prerequisites: ESP32 board (tested: C3, S3, C6; others likely compatible), Wi-Fi connectivity, LLM backend/API key (Anthropic, OpenAI, OpenRouter, Ollama), and a Telegram token/chat ID. Secure flash mode requires --flash-mode secure.
• Setup: bootstrap.sh clones/updates and runs install.sh. Post-flashing, credentials are provisioned via ./scripts/provision.sh or ./scripts/provision-dev.sh. Runtime credentials can be updated without reflashing.
• Links: Full Docs: zclaw.dev, Getting Started: zclaw.dev/getting-started.html.

Highlighted Details

• Chat via Telegram or hosted web relay.
• Timezone-aware scheduling (daily, periodic, one-shot).
• Built-in and user-defined tools, including GPIO read/write control.
• Persistent memory across reboots.
• Persona options: neutral, friendly, technical, witty.
• Provider support: Anthropic, OpenAI, OpenRouter, Ollama.
• Tested hardware: ESP32-C3, ESP32-S3, ESP32-C6.

Maintenance & Community

No specific details on contributors, sponsorships, or community channels (e.g., Discord, Slack) were found in the provided README text.

Licensing & Compatibility

The project is licensed under the MIT license. This permissive license generally allows for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

The primary limitation is the strict hardware dependency on ESP32 microcontrollers. The <= 888 KiB firmware size target is aggressive and includes the entire runtime, networking, and crypto stacks, demanding careful optimization. Initial setup requires provisioning of network and LLM credentials.