Getting Started with Open Claw

Your Open Claw instance is being deployed. Please allow up to 5 minutes for it to become available, then follow the steps below to log in.

Access Your Instance

Click the link below to open your Open Claw instance:

Launch Open Claw

You will be greeted with a login screen. Enter the credentials below to get started.

Username: clawmine

Password: Password

Code Server

A browser-based VS Code editor with a terminal is included with your instance:

Launch Code Server

Password: Password

Essential Chat Commands

These commands are typed directly in the Open Claw chat. Use them to control your session:

Pro tip: If the assistant seems to lose context or exceeded it's context window, you'll want a clean slate, just type /new.

Installing Skills from ClawHub

Open Claw can learn new skills from ClawHub, a marketplace of community and official skills. Use these commands in chat:

Or use direct shell commands via !:
! clawhub search "weather"
! clawhub install weather

Open Claw Has Its Own Tools

Your instance comes with Homebrew and Docker pre-installed. The assistant can:

• Install packages via brew install
• Run Docker containers
• Write and execute scripts
• Create its own skills in ~/.openclaw/skills/

Example: Set Up Privacy-First Web Search

Paste this prompt into your Open Claw chat to set up a local SearXNG search engine (no API keys, no tracking):

Set up a local SearXNG search container you can use for web searches. Run it on port 8888 with Docker (Docker is already installed ready for you to use). Use the searxng/searxng image. Name the container "searx". Then create a skill at ~/.openclaw/skills/searx-search/SKILL.md that teaches you how to query it via curl with format=json.
This gives you a privacy-focused search engine the assistant can use without API keys or rate limits.

Troubleshooting & Restarts

If Open Claw becomes unresponsive or you need to restart the gateway:

Option 1: Use Code Server terminal
1. Open Code Server
2. Open a terminal (Terminal → New Terminal)
3. Run: openclaw gateway --force

Option 2: SSH into your instance
openclaw gateway --force
The --force flag kills any existing gateway process and starts fresh.

Agent Delegation (Subagents)

Open Claw can spawn subagents — specialized worker sessions that run tasks in the background and report back. This lets your main agent delegate research, coding tasks, or long-running jobs without blocking your conversation.

Why Use Subagents?

• Parallel work: Research something while you continue chatting
• Specialized models: Run cheap/fast models for simple tasks, save your main model for complex work
• Isolated context: Each subagent has its own session history

Quick Setup Prompt

Paste this into your Open Claw chat to create specialized subagents ready for delegation:

```
Update the OpenClaw config to enable agent delegation. Add the following to the agents.defaults.subagents section:

1. Set maxSpawnDepth to 2 (allows orchestrator pattern)
2. Set maxConcurrent to 4
3. Set runTimeoutSeconds to 3600 (1 hour timeout)
4. Configure a cheaper default subagent model if available
5. Also create a "research" agent in agents.list with:

id: "research"
subagents.model pointing to a fast/cheap model
subagents.allowAgents: ["*"] (can spawn any agent)
After updating config, show me a summary of what was configured
```

How to Delegate

Once configured, your main agent can delegate naturally:

Subagent Commands

Pro tip: Subagents are best for "fire and forget" tasks. They announce results back to your chat when complete, so you can keep working without waiting.

Other Info

We've tuned a few defaults to make your experience smoother:

• Compaction uses DeepSeek v3.2 — it gave the best results in our tests for summarizing long conversations without losing important context.
• 24-hour agent timeout — your agent won't get cut off mid-task, even on long-running research or coding jobs.
• Homebrew and Docker are ready to go — install packages or spin up containers without any extra setup.

Tweaking the Config

Your Open Claw config lives at ~/.openclaw/openclaw.json. You can edit it directly in Code Server, or just ask your agent to make changes for you:

"Update my config to enable verbose mode by default"

The gateway should auto-restart as soon as it detects config changes. If it doesn't pick up the change, don't be afraid to force it:

openclaw gateway --force