Skip to main content

Deploy on MaixCam / MaixCam2

This document walks you through integrating Sipeed MaixCAM or MaixCAM2 AI cameras with PicoClaw.

MaixCAM and MaixCAM2 are hardware platforms designed for real-world AI vision, audio, and AIoT applications. They provide cost-effective, high-performance processors, along with cameras, displays, Wi-Fi, and an easy-to-use software ecosystem. By deploying PicoClaw on MaixCam or MaixCAM2, you can enable scenarios such as smart monitoring and edge inference.

Hardware​

  • MaixCAM
  • MaixCAM-Pro
  • MaixCAM2

Prerequisites​

  • The firmware has been flashed to your MaixCam or MaixCAM2 device, and the device can boot and connect to the network.

  • You can access the device through MaixVision or SSH.

Deployment​

Configure Network​

  1. Configure networking: if your device is not connected yet, follow MaixPy network setup first.
  1. Get the device IP: go to Settings -> Device Info -> IP Address, or find it from the MaixVision device list on the same LAN.

Connect via SSH and Deploy PicoClaw​

  1. Make sure your computer and MaixCam/MaixCam2 are on the same LAN.
  2. Use an SSH client to connect to the device IP. The default username/password is root / root.
  3. In the SSH session, run the commands below to download and start PicoClaw:
curl -L# -o picoclaw_Linux_riscv64.tar.gz \
https://github.com/sipeed/picoclaw/releases/latest/download/picoclaw_Linux_riscv64.tar.gz

mkdir -p /root/picoclaw
gzip -dc picoclaw_Linux_riscv64.tar.gz | tar -xvf - -C /root/picoclaw
cp /root/picoclaw/picoclaw* /usr/bin
rm -rf /root/picoclaw /root/picoclaw_Linux_riscv64.tar.gz

picoclaw-launcher-tui

You can now use PicoClaw TUI in the terminal for configuration and management.

  1. To start PicoClaw Web UI, run:
picoclaw-launcher -no-browser -public

Then open http://<device-ip>:18800 in a browser on the same LAN.

Connect via MaixVision and Deploy PicoClaw​

  1. Connect to your MaixCam or MaixCam2 device in MaixVision.

  2. In MaixVision, create a new Python file and paste the script content below:

    Script: install_picoclaw.py

  3. Click Run. PicoClaw will be downloaded and installed on the device, and the PicoClaw Web UI will start.

  4. Open http://<device-ip>:18800 in a browser on the same LAN to access the PicoClaw Web UI.

Usage example​

Below is a simple hands-on example showing how to use PicoClaw with MaixCAM for AI vision development:

  1. Basic setup: First configure the model and chat channel in the Web UI or TUI. The example below uses Discord and the glm-4.7 model:

  2. Generate capture script: In Discord, talk to PicoClaw and ask it to write and run a MaixPy capture script:

    Refer to https://wiki.sipeed.com/maixpy/doc/zh/vision/camera.html Please write a MaixPy script that captures an image and saves it to /root/capture.jpg

    capture

  3. Generate an image-analysis Skill: Ask PicoClaw to automatically generate a Skill that analyzes the image just captured using a large model:

    Can you write a skill that uses a large model to analyze the image we just captured?

    image_analysis_skill

  4. Run image analysis: Invoke the skill you created to perform an in-depth analysis of the image:

    Use the skill created earlier to analyze this image. You can use the glm-4.6v model; the API_KEY can be found in .picoclaw/config.json.

    image_analysis_result

FAQ​

  • Large models may not complete complex tasks perfectly in a single turn. If you encounter errors or deviations, guide the model through additional conversational turns to correct them.
  • This is a basic image-text workflow example. PicoClaw can do much more—customize models and automated workflows to fit your use case.

References​

For more MaixCam hardware information and usage guides, see: