Back to Claws
PicoClaw logo
ClawDeviceNeeds review

PicoClaw

Ultra-lightweight personal AI assistant written in Go that runs on $10 hardware with under 10MB RAM and boots in under 1 second, delivering OpenClaw-like agent capabilities on embedded devices, Raspberry Pi, and even old Android phones.

Decision information

Manual verification pending
Ecosystem category
Claw Agents
Agent class
Device
Lineage
Independent
Runs
Device
Autonomy
Autonomous
Content type
Product
Pricing
Open Source
Open source
Yes
License
Not recorded
Maintenance
Unknown
Install methods
Source
Platforms
Not recorded
Interfaces
Telegram, Discord, Terminal

Documented Claw signals

View inclusion standard
Tool useMulti-step goalsAutonomous executionBrowser / shell / computer access

Security signals (informational, not a security certification)

Source available

Overview

PicoClaw is an ultra-lightweight personal AI assistant inspired by nanobot and refactored from the ground up in Go through a self-bootstrapping process (95% AI-generated core). Developed by Sipeed, it brings powerful agent capabilities to extremely constrained hardware while maintaining full functionality for personal automation.

It runs on any Linux board as cheap as $10 (e.g., LicheeRV-Nano, Raspberry Pi Zero) with a memory footprint under 10MB and startup under 1 second — 99% smaller and 400× faster than comparable tools like OpenClaw.

Official website: picoclaw.io

Features

  • Ultra-Lightweight: <10MB RAM footprint — 99% smaller than OpenClaw
  • Blazing Fast: Boots in <1s even on 0.6–0.8 GHz single-core hardware
  • Single Binary: Zero dependencies, cross-platform (RISC-V, ARM, MIPS, x86_64, ARMv7)
  • Multi-LLM Support: Cloud providers (OpenAI, Anthropic, DeepSeek, GLM, Ollama local, etc.) plus model routing
  • Rich Integrations: Telegram, Discord, WhatsApp (native), Matrix, QQ, DingTalk, WeCom, LINE, Slack, IRC and more via shared gateway
  • Agent Capabilities: Tools, memory, cron jobs, skills, multi-agent orchestration, workspace sandboxing
  • Launcher & Web UI: Optional web console and TUI for management
  • Security-First: Sandboxed execution, dangerous command blocking, workspace restrictions

Performance Comparison

AspectOpenClawNanoBotPicoClaw
LanguageTypeScriptPythonGo
RAM>1GB>100MB<10MB
Startup (0.8GHz)>500s>30s<1s
Hardware CostMac Mini $599SBC ~$50As low as $10

Use Cases

  • 24/7 personal AI on Raspberry Pi Zero or $10 RISC-V boards
  • Always-on Telegram/Discord/WhatsApp bot for reminders, email, automation
  • Edge AI on old Android phones via Termux or embedded Linux devices
  • Lightweight alternative to heavy agents for home servers, IoT, or resource-limited environments
  • Self-hosted automation without GPUs or cloud bills

Quick Installation

Binary (recommended) Download latest release from GitHub Releases for your architecture.

One-command onboard

picoclaw onboard

Configure Edit ~/.picoclaw/config.json with your LLM API keys and preferred channels.

Run

picoclaw agent          # direct chat
picoclaw gateway        # messaging bots
picoclaw launcher       # web console at http://localhost:18800

Docker support and full source build instructions available in README.

Supported Platforms

  • Raspberry Pi (including Zero 2 W)
  • Sipeed LicheeRV-Nano, MaixCAM, NanoKVM
  • Any Linux (x86, ARM, RISC-V, MIPS, LoongArch)
  • Android via Termux
  • macOS, Windows (experimental)

Community & Development

Actively maintained with frequent releases (latest v0.2.2 as of March 2026). MIT licensed, 140+ contributors. Roadmaps, contributing guide, and skill ecosystem on GitHub.

PicoClaw proves AI agents can be truly ubiquitous — deploy anywhere, even on the smallest hardware.

Tags

aiagentlightweightgoembeddedraspberry-pirisc-vopen-sourceautomationtelegramdiscordcli