picoclaw
Ultra-lightweight AI assistant for resource-constrained devices
TLDR
Initialize configuration and set up credentials
SYNOPSIS
picoclaw [command] [options]
DESCRIPTION
picoclaw is an ultra-lightweight personal AI assistant written in Go, designed to run on extremely resource-constrained hardware. It uses less than 10MB of RAM and ships as a single self-contained binary with sub-second boot times.
PicoClaw supports RISC-V, ARM64, and x86 architectures natively, making it deployable on devices ranging from $10 RISC-V boards to powerful servers. Despite the minimal footprint, it provides full agent capabilities including planning, web search integration, command execution, and automation workflows.
The tool supports multiple LLM providers including OpenAI, Anthropic (Claude), Google Gemini, and Zhipu. It can integrate with chat platforms via the gateway command for bot deployments on Telegram, Discord, and other messaging services.
PARAMETERS
onboard
Initialize configuration and set up the environment for first-time use. Creates config at ~/.picoclaw/config.json.agent
Start an interactive CLI chat session with the AI agent.agent -m message
Send a single message to the agent and receive a response without entering interactive mode.gateway
Start the multi-channel gateway for chat platform integrations including Telegram, Discord, QQ, DingTalk, LINE, and WeCom.
CONFIGURATION
Configuration is stored in ~/.picoclaw/config.json. Run picoclaw onboard for interactive setup.
Key configuration fields:
"agents": {
"defaults": {
"workspace": "/path/to/workspace",
"model_name": "claude-opus-4-5",
"max_tokens": 8192,
"temperature": 0.7,
"max_tool_iterations": 10
}
},
"model_list": [
{
"model_name": "claude",
"model": "claude-opus-4-5",
"api_key": "sk-ant-xxx"
}
]
}
CAVEATS
PicoClaw executes commands on your machine with your user permissions. API keys are stored in plaintext in the config file. The project was largely AI-bootstrapped with approximately 95% of the core system generated by an AI agent, which may affect code quality and maintainability. Being Go-based, it requires downloading precompiled binaries or building from source rather than installing via a package manager.
HISTORY
PicoClaw was developed by Sipeed, a company known for RISC-V hardware, as an AI assistant optimized for their low-cost development boards. First released in 2025, it was written in Go through a self-bootstrapping process where the AI agent itself drove the architectural design and code optimization. The project gained attention for running full agent capabilities on hardware costing as little as $10.

