Quick Start

This guide helps you quickly complete the first deployment of Ameba-Claw and experience the complete workflow of running an AI Agent on an Ameba development board.

Prerequisites

Before you begin, confirm you have the following ready.

Hardware

  • Realtek Ameba development board: RTL8721F

  • USB data cable: for connecting the board to your PC (ensure the cable supports data

    transfer, not charging only)

  • Wi-Fi network: 2.4 GHz band with internet access

LLM API Key

Ameba-Claw supports any OpenAI-compatible or Anthropic-compatible LLM service. The following are common providers where you can apply for an API key:

After signing up, note down the following two pieces of information for the LLM configuration step:

  • API Key: the authentication credential issued by the provider

  • Model ID: the model name, e.g. deepseek-v4-flash

Messaging Bot (Optional)

  • WeChat: To chat with the Agent through WeChat, simply scan a QR code in the web

    console to bind your account.

For Feishu, QQ, or Telegram, you will need to apply for an AppID & AppSecret or Token on the respective open platform.

Tip

If you just want a quick first look, you can skip this step. The built-in WebIM in the web console lets you chat with the LLM directly.

Step 1: Flash Firmware & Wi-Fi Setup

The Ameba-Claw firmware is flashed to the board through an online flash page — no software installation required. Wi-Fi is then configured by sending an AT command over the serial console.

Note

This section provides a quick overview. For the full guide, see Flashing Firmware and Wi-Fi Configuration.

Flashing and Setup Steps

  1. Open the online flash page

    Open the Ameba-Claw online flash tool in Chrome or Edge.

  2. Connect the board

    Plug the USB data cable into the board’s USB port and connect the other end to your PC.

  3. Select firmware and device

    In the “Select application” section, choose the platform and application firmware. In the “Select device” section, select your board model.

  4. Click Flash

    Click “Start flashing”. Select the corresponding COM port from the serial port dialog that appears and confirm. The tool flashes the firmware automatically — approximately 1–2 minutes. Do not disconnect or close the page.

  5. Connect the console

    After flashing completes, scroll to the “Console” section at the bottom of the page. Click “Connect”, then select the COM port and confirm.

  6. Send the Wi-Fi command

    Once the console shows “Connected”, type the following in the input box — replacing <SSID> and <PASSWORD> with your network name and password — and press Enter:

    AT+WLCONN=ssid,<SSID>,pw,<PASSWORD>

  7. Confirm the connection

    When the console log shows the following, the board is connected to Wi-Fi:

    STA connected (DHCP), IP=<device IP>

    Enter the device IP in your browser address bar to open the web console.

Tip

Note down the board’s IP address — you can access the web console at any time by entering it directly in your browser’s address bar.

Step 2: Configure LLM

Enter the board’s IP address in the browser address bar to open the web console.

Note

For the full web console documentation, see AmebaClaw Web Console User Manual.

Configuration Steps

  1. Open the web console

    From a device on the same Wi-Fi network as the board, navigate to http://<device IP>.

  2. Go to LLM Config

    Click LLM Config in the left navigation bar.

  3. Use a provider preset

    The top of the page provides preset templates for common providers. Click a provider icon (e.g. DeepSeek, OpenAI) to automatically fill in the corresponding API URL and recommended model name.

  4. Enter the API Key

    Paste your key into the API Key field.

  5. Save the configuration

Step 3: Start Chatting

Once configuration is complete, you can talk to the Agent in the following ways.

Method 2: WeChat Bot

WeChat requires no token application — simply scan a QR code to bind, making it the easiest way to connect a messaging app:

  1. In the web console, click IMBOT Config → WeChat → Generate QR Code.

  2. The page displays a QR code — scan it with WeChat to complete the binding.

  3. Once bound, send a message to the bot to start chatting.

First Conversation Suggestions

Try the following messages to get a feel for the Agent’s basic capabilities:

  • Introduce yourself

  • What time is it now

  • What can you do

  • Help me list today's to-do list

Tip

The Agent supports multi-turn conversations and has long-term memory. Information you share with it (such as your name or preferred language) is stored in vfs:/USER.md and persists across conversations even after a reboot.

Next Steps

Congratulations on completing your first Ameba-Claw deployment! The following resources can help you explore further: