ClawPanel Documentation

ClawPanel is a visual management panel for multiple AI Agent frameworks, specifically supporting OpenClaw and Hermes Agent. It features an integrated AI assistant to help with installation, configuration diagnostics, and troubleshooting.

Key capabilities include:

• Multi-Engine Architecture: Independent management of OpenClaw and Hermes Agent.
• Hermes Agent Chat: Visualized tool calls, file access, and SSE streaming.
• Service Management: Start/stop gateways, version detection, and upgrades.
• Model Configuration: Multi-provider management with connectivity testing.
• Automation: Cron-based scheduled tasks and webhook support.
• Communication: Unified management for Telegram, Discord, Feishu, DingTalk, and QQ.

The following projects are part of the ecosystem and may be useful for your development:

• OpenClaw: An AI agent framework.
• ClawApp: A cross-platform mobile chat application.
• cftunnel: A Cloudflare Tunnel tool.

ClawPanel uses a multi-engine architecture:

• OpenClaw: Best for Agent management and Gateway operations for users already in the OpenClaw ecosystem.
• Hermes Agent: Focused on the long-term operational experience of sessions, memory, personality, and tool execution.

Both engines can be managed unified within the same ClawPanel interface.

Follow these steps to get your AI Agent environment running:

1. Initial Setup: Run the application; it will automatically detect Node.js, Git, and OpenClaw. Follow one-click installation prompts if components are missing.
2. Configure Models: Add AI providers (e.g., DeepSeek, OpenAI, Ollama) and use the connection test feature.
3. Start Gateway: Navigate to Service Management (Quản lý dịch vụ) and click "Start" (Khởi động). A green status indicates it is ready.
4. Chat: Go to the Direct Chat (Trò chuyện trực tiếp) interface, select your model, and begin interacting.

For ARM64 devices, you can use the official Docker image which includes both OpenClaw and ClawPanel.

One-click Docker Start

docker run -d \
  --name openclaw \
  -p 1420:1420 \
  -p 18789:18789 \
  -v openclaw-data:/root/.openclaw \
  --restart unless-stopped \
  ghcr.io/qingchencloud/openclaw:latest

If pulling from GHCR is slow, use the Tencent Cloud mirror:

docker run -d \
  --name openclaw \
  -p 1420:1420 \
  -p 18789:18789 \
  -v openclaw-data:/root/.openclaw \
  --restart unless-stopped \
  ccr.ccs.tencentyun.com/qingchencloud/openclaw:latest

To effectively deploy and iterate on a Hermes Agent, follow these steps:

1. Configure Model and Gateway: Ensure the Hermes Agent can connect to model services.
2. Initialize Long-term Memory: Populate the Notes, User Profile, and Soul in the Agent Memory page.
3. Verify via Sessions: Engage in dialogue to confirm that personality, preferences, and context are working as expected.
4. Connect Messaging Channels: Integrate external channels such as QQ, Telegram, or Discord based on your use case.
5. Iterate Memory Assets: Periodically organize facts, preferences, and rules gathered from real usage back into the long-term memory files.

For development boards like Orange Pi, Raspberry Pi, or RK3588, you can use the following methods:

1. Docker ARM64: Run the official image: docker run ghcr.io/qingchencloud/openclaw:latest
2. Manual Execution: Use npm run serve to run the web service.

Requirements:

• Node.js 18+ for the Web version.
• OpenClaw Gateway recommends Node.js 22.19.0+.

Prerequisites

• Node.js >= 18
• Rust (stable)
• Tauri v2 system dependencies

Installation

git clone https://github.com/qingchencloud/clawpanel.git
cd clawpanel
npm install

Development (macOS / Linux)

# Start full Tauri desktop application
./scripts/dev.sh

# Start Vite frontend only (browser debugging with mock data)
./scripts/dev.sh web

Development (Windows)

# Start full Tauri desktop application
npm run tauri dev

# Start Vite frontend only (browser debugging with mock data)
npm run dev

The ClawPanel AI assistant operates in four distinct modes that determine its system permissions and interaction style. The UI color changes based on the active mode to provide clear visual feedback on current permissions.

Note: Tool switches (Terminal/File) in Settings take precedence over modes. If the Terminal is disabled, the AI cannot execute commands even in Infinite mode.

Ciyapi provides OpenAI-compatible access to mainstream models (GPT, Claude, Gemini, etc.). To integrate it into ClawPanel:

1. Register and obtain an API Key at ciyapi.79tian.com.
2. In ClawPanel, add a new AI provider.
3. Set the OpenAI-compatible Base URL to: https://ciyapi.79tian.com/v1.
4. Use your obtained API Key.