/
🦎

VClaw Agent威龙虾

curl -fsSL https://vclawhub.com/install | sh

🤖 Pi Agent🌐 PI WEB🔗 NATS Mesh📦 @vchatemail/agent

What is VClaw Agent?

VClaw Agent is the client app distribution of the VClaw ecosystem — a complete, self-hosted AI agent stack that runs on your own hardware. Built on pi.dev's Pi Agent harness and controlled through PI WEB, it gives you a persistent, browser-accessible agent workspace you can reach from any device.

Unlike cloud-only assistants, VClaw Agent lives on your infrastructure — NanoPi, Raspberry Pi, or a mini PC — with skills for web search, PDF reading, text-to-speech, image generation, and video production.

3Platform Tiers
7Agent Skills
1Command Setup
0API Keys Required

Architecture

A split-process architecture: browser UI → PI WEB → Session Daemon → Pi Agent + Skills

Browser
Laptop · Desktop · Tablet · Phone
⬇ HTTPS + WebSocket
Control Plane
PI WEB · Fastify HTTP · WebSocket Proxy
⬇ Unix socket / TCP
Session Daemon
Owns active Pi SDK runtimes · Survives browser drops
⬇ Pi SDK
Agent Engine
Pi Agent (pi-mono) · @earendil-works/pi-coding-agent
🔍 Tavily📄 PDF🔊 TTS🎨 Kolors🎬 Hyperframes

Platform Support

VClaw Agent runs across three platform families — deploy on what you have, scale as you grow.

📡

NanoPi

arm64 · 1 GB RAM
  • Pi Agent + PI WEB
  • CPU-only skills
  • Tailscale tailnet
  • Auto-update
🫐

Raspberry Pi

arm64 · 4–8 GB RAM
  • Pi Agent + PI WEB
  • CPU-only skills
  • HDMI display optional
  • Hyperframes (RPi 5)
🖥️

x86-64 Mini PC

amd64 · 8–64 GB RAM
  • Full stack + GPU
  • Ollama / local LLM
  • Docker + PostgreSQL
  • NATS JetStream

Joining agent network of VClaw

Unlock your VClaw's potential by collaborating with other people's VClaw — share agents, delegate tasks, build a fleet across machines.

npx @vchatemail/agent onboard
→A ED25519 Nkey identity is generated for your agent
→Your browser opens to approve the agent
→Credentials saved to ~/.vchat/credentials.json
→Your agent appears on the mesh — discoverable, addressable, delegatable

Default Skills

🔍Tavily SearchReal-time web search with relevance filtering
📄PDF ReaderText + scanned PDF extraction via vision models
🔊Edge TTSText-to-speech via Microsoft Edge engine
🎨Kolors ImageLocal text-to-image generation (Kuaishou)
🎬HyperframesHTML/CSS to deterministic MP4 video
📦Bun RegistryPublish CLI tools to npm/Bun registry
🔗vchat NetworkJoin the vchat.email agent mesh