PicoClaw
Ultra-lightweight personal AI assistant. Memory under 10MB, starts in under one second—from $10 hardware to servers. AI within reach, everywhere.
No heavy runtime. Single binary. One command to run.
Why PicoClaw?
Rewritten in Go for a 10× performance boost. From Raspberry Pi to cloud, from CLI to chat apps—run your AI assistant anywhere.
Ultra-Low Resource Usage
Memory under 10MB—about 99% less than alternatives. Runs smoothly on $10 hardware and scales to servers. No bloat, no long installs.
Lightning-Fast Startup
Starts in under one second, even on single-core 0.6GHz. Say goodbye to long waits—start conversations or run picoclaw agent instantly.
Multi-Architecture
x86_64, ARM64, and RISC-V. One binary, one click. Deploy on Raspberry Pi, Mac, Linux, Windows, or Docker. See comparison with OpenClaw and NanoBot.
Get running in one minute
Clone, configure once, then chat from the CLI or your favorite messenger.
$ git clone https://github.com/sipeed/picoclaw && cd picoclaw && picoclaw onboard
$ picoclaw agent -m "Hello, introduce yourself"Full setup: Get Started · Docker · Configuration
Guides & use cases
Step-by-step guides: Raspberry Pi & Pi 5, VPS, systemd, nginx, Cloudflare Tunnel, Tailscale, GitHub Actions, Home Assistant, Docker, n8n, OpenAI GPT, Claude, Gemini, Groq, DeepSeek, local APIs, and cron.
- All guides
- Blog
- Raspberry Pi AI assistant
- Self-hosted AI assistant
- Home Assistant + AI
- Docker homelab AI
- Linux cron AI jobs
- Windows WSL2 AI assistant
- Telegram bot setup
- Discord bot setup
- Local LLM with Ollama
- Raspberry Pi performance tuning
- RISC-V setup
- OpenRouter vs OpenAI
- Docker Compose for PicoClaw
- Groq + PicoClaw
- DeepSeek API + PicoClaw
- OpenAI-compatible local API
- n8n webhooks
- systemd service (Linux)
- VPS headless AI agent
- Gemini + PicoClaw
- Claude + PicoClaw
- OpenAI GPT + PicoClaw
- nginx HTTPS proxy
- Cloudflare Tunnel
- Tailscale homelab
- Raspberry Pi 5 assistant
- GitHub Actions
- Use cases
- Comparison
- Lightweight AI alternatives
How it works
Three steps from zero to a working AI assistant—locally or on your favorite platform.
Clone & install
Download the binary or clone the repo. Run picoclaw onboard to create your config and workspace.
Add your API key
Set your LLM key in ~/.picoclaw/config.json. Use OpenRouter, Zhipu, OpenAI, or others—many have free tiers.
Chat or connect
Run picoclaw agent for CLI chat, or picoclaw gateway and connect Telegram, Discord, or more.
Built for real-world use
From personal productivity to professional workflows. Your data stays local; you choose the models and platforms.
Code assistant
Local code completion, debugging, and documentation. Keep proprietary code private while getting AI help. Works with your workspace and sandbox.
Content creation
Draft blog posts, emails, and documents with AI. Everything stays on your device—privacy first. No data sent to third-party clouds except your chosen LLM API.
Personal assistant
Scheduled tasks, reminders, and automation. Run Telegram or Discord bots with picoclaw gateway. One config, many channels.
Multi-model support
Use OpenRouter, Zhipu, Anthropic, OpenAI, DeepSeek, Groq, and more. Switch models in config—find the best mix of speed, cost, and quality.
“Ultra-lightweight personal AI assistant. Memory under 10MB, starts in under 1 second. From $10 hardware to servers—AI within reach.”
Explore
Docs, integrations, and everything you need to run PicoClaw your way.
Product
Features · Comparison · Download · Docs
Get going
Get Started · Configuration · Docker · API
Learn & support
Use Cases · FAQ · Community · Contribute
Official: picoclaw.net · GitHub