OpenClaw is an open-source AI Agent framework that can connect to local or cloud Large Language Models (LLMs) to achieve automated tasks and intelligent interaction. This document will introduce how to perform image flashing, initial configuration, and initial access for OpenClaw on AI Pyramid.
Download the OpenClaw image file below and refer to AI Pyramid Image Burning to complete the burning.
| Firmware Version | Download Link |
|---|---|
| AI_Pyramid_openclaw_emmc_ubuntu_rootfs_desktop_V3.6.4 | Download |
After completing the image burning, the device will automatically enable the OpenClaw initialization boot page (only displayed when the initialization is not completed).
https://pyramid-openclaw.local through a browser.Get Started and click Open Web Terminal to enter the terminal.
Connect a monitor, mouse, and keyboard to the AI Pyramid. In the desktop browser, access https://pyramid-openclaw.local. The operation method is the same as LAN access.
In the Web Terminal, enter the username root and password 123456 to log into the device terminal:
Take the QuickStart configuration method to access the Qwen API as an example. Directly run the following commands to start the initialization (the device is pre - installed with the OpenClaw service):
openclaw onboard Tip: Initialization usually takes about 2 minutes. Please be patient.
Yes to continue:◆ I understand this is personal - by - default and shared/multi - user use requires lock - down. Continue?
● Yes / ○ No QuickStart:◆ Onboarding mode
│ ● QuickStart (Configure details later via openclaw configure.)
│ ○ Manual In this case, Qwen (OAuth) is used as an example. You can also choose other providers according to your subscription:
◆ Model/auth provider
│ ○ OpenAI
│ ○ Anthropic
│ ● Qwen (OAuth)
│ ○ OpenRouter
│ ○ Skip for now
◇ Qwen OAuth
│ Open https://chat.qwen.ai/authorize?user_code = 1 - _VFKMX&client = qwen - code to approve access.
│ If prompted, enter the code 1 - _VFKMX.
Keep current to keep the default model:◆ Default model
│ ● Keep current (qwen - portal/coder - model)
│ ○ Enter model manually
Skip for now to skip the communication channel configuration:◆ Select channel (QuickStart)
│ ○ Telegram (Bot API)
│ ○ Discord (Bot API)
│ ● Skip for now No to not install Skills for now:◆ Configure skills now? (recommended)
│ ○ Yes / ● No Skip for now and press Enter to confirm:◆ Enable hooks?
│ ◼ Skip for now
│ ◻ 🚀 boot - md
│ ◻ 📝 command - logger token value. This is the credential for accessing the OpenClaw console later:◇ Dashboard ready
│ Dashboard link (with token):
│ http://127.0.0.1:18789/#token = 0c4a936a196d3be5812e7624d5cb419b0b364c308f68b799
│ Copy/paste this URL in a browser on this machine to control OpenClaw.
openclaw config set gateway.bind lan
openclaw config set gateway.controlUi.allowInsecureAuth true
openclaw config set gateway.controlUi.allowedOrigins '["http://127.0.0.1","http://127.0.0.1:18789","https://pyramid-openclaw.local"]'
openclaw config set gateway.controlUi.enabled true
YOUR_TOKEN_VALUE with the token saved in step 10):Within the same local area network, access through the following domain name plus the token parameter:
https://pyramid-openclaw.local#token = YOUR_TOKEN_VALUE openclaw devices list Then approve the pairing request (replace <requestId> with the actual device ID):
Note: The requestId is only valid for the current browser. If you change the device or browser, you need to view the requestId again and re - initiate the pairing.
openclaw devices approve <requestId>
After the configuration is completed, the console can be accessed normally on the web - side.
Submit the following task in the OpenClaw console to verify the deployment effect:
Generate a complete Tetris web mini - game, save it as openclaw - deploy - game.html, then start a local HTTP server in the directory where the file is located, and return the accessible URL address.
After OpenClaw completes the task, copy the returned URL and open it in a browser to view the mini - game effect.