Getting Started
LiberClaw lets you create and deploy AI agents on decentralized infrastructure. This guide walks you through signing up, creating your first agent, and sending your first message.
Sign up
Section titled “Sign up”LiberClaw supports multiple sign-in methods depending on your platform:
Web (app.liberclaw.ai):
- Email magic link — Enter your email and click the link sent to your inbox
- Google OAuth — One-click sign-in with your Google account
- GitHub OAuth — Sign in with your GitHub account
Mobile (iOS/Android):
- Email magic link — Same as web
- Native Google sign-in (Android) or Apple sign-in (iOS)
- Guest mode — Start immediately without an account (limited session and memory)
Guest accounts have limited memory and session limits. You can upgrade to a full account later by signing in with email or OAuth.
Create your first agent
Section titled “Create your first agent”Once signed in, you land on the agents dashboard. Tap Create Agent to start the five-step wizard:
Step 1: Identity
Section titled “Step 1: Identity”Choose a name for your agent (up to 50 characters). You can also optionally link a Telegram bot at this stage by providing a bot token from @BotFather.
Step 2: Soul (System prompt)
Section titled “Step 2: Soul (System prompt)”Define your agent’s personality and behavior by writing a system prompt, or pick a template to pre-fill the prompt, model, and skills automatically. Available templates are shown as selectable pills — tapping one loads its configuration, and tapping it again deselects it.
Step 3: Skills
Section titled “Step 3: Skills”Choose optional skills that teach your agent specific workflows. Skills are organized into categories (developer, productivity, web3). If you skip this step, your agent uses its default capabilities (bash terminal, file system access, web tools).
Step 4: Model
Section titled “Step 4: Model”Select the inference model your agent will use:
| Model ID | Brand Name | Description |
|---|---|---|
qwen3-coder-next | Claw-Core | Default model, optimized for coding tasks |
glm-4.7 | Deep-Claw | Alternative model |
Step 5: Review & Deploy
Section titled “Step 5: Review & Deploy”Review your configuration (name, model, system prompt, skills) and hit Deploy. You are redirected to the agent detail page where you can watch the deployment progress in real time.
Deployment
Section titled “Deployment”Deployment provisions a VM on Aleph Cloud and sets up your agent. The process has six steps:
- Infrastructure Provisioning — Allocating compute on Aleph Cloud
- Network Allocation — Waiting for the VM to come online
- Secure Connection — Establishing SSH connection to the VM
- Environment Setup — Installing the runtime and deploying code
- Service Activation — Starting the agent and configuring HTTPS
- Health Check — Verifying the agent is responding
Deployment typically takes 2-3 minutes. A circular progress indicator and terminal log show real-time status. If deployment fails, you can retry with the Rebuild button.
Send your first message
Section titled “Send your first message”Once deployment completes and the health check passes, the agent detail page shows a green “Agent is healthy and responding” status. Click the Chat button to open the chat interface and start talking to your agent.
Type a message in the input field and press send. The agent responds via server-sent events (SSE), streaming text in real time. You will see tool use indicators when the agent executes tools like bash commands or file operations.