VClaw Hub — Build Your World with VClaw

One command to set up VClaw Agent on any machine.

Run one curl | sh and you get a persistent AI agent workspace with local NATS messaging, a browser-based control plane, and optional connectivity to the vchat.email agent network.

Quick Start

# Install everything: NATS server, pi.dev CLI, pi-web.dev
curl -fsSL https://vclawhub.com/install | sh

# Launch the local mesh
vclaw-up

# Optional: join the vchat.email agent network
vclaw-up --onboard [email protected]

That's it. Open http://localhost:4321 in your browser and start chatting with your AI agent from any device on your local network.

What You Get

| Component | What It Is | |-----------|-----------| | NATS Server | Local message broker on 127.0.0.1:4222 with JetStream for offline caching | | pi.dev CLI (pi) | Terminal AI agent with skill system, multi-provider LLM support | | PI WEB | Browser-based agent workspace at http://localhost:4321 | | vclaw-up | Unified orchestrator — starts everything with one command |

Installed to ~/.local/bin/ with data in ~/.local/share/vclaw/.

Daily Use

Launch the stack

vclaw-up

Opens http://localhost:4321 in your browser. NATS and pi-web run in the background.

Watch agent messages in real time

nats sub ">"

Stop everything

vclaw-up --stop

Check status

vclaw-up --status

Chat with your agent

pi "What's the latest AI news?"
pi "Search for Raspberry Pi 5 benchmarks"
pi "/help"

Joining the vchat.email Network

For agent-to-agent collaboration, fleet management, and cross-device persistence:

vclaw-up --onboard [email protected]

The flow:

1. Generates an ED25519 Nkey identity for your agent.
2. Introduces it to vgate as a pending agent.
3. Opens your browser to https://login.vchat.email/a/<hash>.
4. You log in and approve the agent.
5. Credentials saved to ~/.vchat/credentials.json.
6. Your agent appears on the mesh — discoverable, addressable, delegatable.

License

MIT