Index / zclaw · updated Oct 24, 2026

zclaw

tnm/zclaw · healthy · rank 26 of 50 by stars

An ultra-minimalist AI personal assistant that squeezes onto an ESP32 microcontroller with an all-in firmware budget of just 888 KiB. It brings GPIO control, cron scheduling, persistent memory, and custom tools to bare-metal hardware.

#esp32#embedded#firmware#tinyml#personal-assistant
Compare vs OpenClaw
GitHub ↗

Facts

Repository

Contributors4
Open issues11
Last commitMay 17, 2026
Release cadence~7 days
Latest releasev2.13.0

Runtime

LanguageC
Memory1 MB
Boot time8 ms
Deploymentedge · self-hosted
Setup difficultyMedium
Plugin ecosystemLimited

Posture

LicenseMIT
Local-firstYes
Cloud dependencyOptional
Multi-userNo
Privacy postureStrong

Community

Sentiment55% positive
Reddit mentions0
Web results10

Security breakdown

Composite 72 / 100 · how these are scored

Sandboxing6 / 10

higher is safer

API security7 / 10

higher is safer

Network isolation5 / 10

higher is safer

Telemetry safety9 / 10

higher is safer

Shell access risk3 / 10

higher is riskier

Model access

Read from the repository, not written by a model · 57 files examined

2 providers · runs locally · gateway support

Direct
AnthropicOpenAI
Gateway
OpenRouter
Local
Ollama
Compatible
OpenAI-compatible

Pinned models

gpt-5.4OpenAI

claude-sonnet-4-6 released 2026-02-17 — 8 months old (from the public model catalogue)

Pin last edited 7 months ago

Several defaults are pinned; the project picks one per provider rather than shipping a single default.

Evidence

Decision

Why choose zclaw over OpenClaw?

Why choose this

  • Runs on $5 ESP32 hardware with 888 KiB total firmware
  • Direct GPIO, I2C, and DHT sensor control built in
  • No server or PC required; fully self-contained device

Tradeoffs

  • Far smaller tool/skill ecosystem than OpenClaw
  • Limited to Telegram or web relay chat channels
  • Constrained by microcontroller RAM and single-device scope

Best fit

  • Makers wanting an AI assistant on ESP32 hardware
  • GPIO and sensor automation via natural language
  • Ultra-low-power, always-on edge deployments

Avoid if

  • You need a rich desktop or server agent runtime
  • You want a large plugin/skill ecosystem
  • You need multi-user or team workflows
Good Confidence70%

Strong evidence from README, docs, and recent commits about architecture and features, but zero Reddit mentions and web results that mostly reference OpenClaw generally rather than zclaw specifically. Community sentiment is therefore estimated with low confidence.

AI layer reviewed Oct 5, 2026 · how this is written

Star activity

2,239 stars today

Overview

zclaw is a C-based AI personal assistant designed to run entirely on ESP32 microcontrollers, with a strict all-in firmware budget of 888 KiB that includes the ESP-IDF/FreeRTOS runtime, Wi-Fi, TLS, and certificate bundles. Despite the tiny footprint, it supports timezone-aware cron schedules (daily, periodic, one-shot), persistent memory, GPIO control, and user-defined tools composable through natural language. Chat happens over Telegram or a hosted web relay, and credentials (Wi-Fi, LLM API keys, Telegram allowlists) can be provisioned or re-provisioned without reflashing, with an optional secure flash mode for encrypted storage.

Recent development is active and hardware-focused: the v2.13.0 release added DHT sensor and generic I2C tools, hardened tool schemas, and trimmed firmware size to stay within budget. The project supports both cloud LLM backends and local models via Ollama, with compile-time rate limits (100/hour, 1000/day defaults) to control API spend.

Compared to OpenClaw, zclaw trades ecosystem breadth and desktop-class capabilities for extreme minimalism and physical-world access. It is the OpenClaw-adjacent option for makers who want an assistant that can toggle pins, read sensors, and run scheduled automations on a $5 board rather than a full computer.

Nominate a clone

Add a new Claw

Paste a GitHub repository and tell us why it belongs on the tracker.

Opens a prefilled issue on GitHub — every nomination is public. Comfortable with a PR? Adding the repo to projects.json is faster.