UNIHIKER K10 + ESP-Claw: Build a Perceptive and Proactively Responsive AI Desktop Butler

Reference Article：ESP-Claw AI Agent Deployment on UNIHIKER K10 (ESP32-S3)

Project Introduction

Most conventional smart hardware merely follows preset commands. Functions such as fan control, light switching and environmental sensing require tedious programming and manual instruction; otherwise, devices stay passively on standby. While they can execute basic tasks, they suffer from slow response and offer limited privacy protection and autonomous decision support.

ESP-Claw fundamentally changes this paradigm. It empowers hardware with local AI without complicated programming, enabling autonomous environmental perception, proactive decision-making and multimodal human-computer interaction. Powered by the ESP32-S3 chip and abundant sensor interfaces of UNIHIKER K10, a thoughtful desktop AI assistant named Jarvis is built with real independent thinking capability.

Jarvis real-time monitors temperature, humidity, ambient light, posture and other multimodal data, feeding back via screen text, RGB lighting and fan operation. Users can control fan switches and speed through voice or text input. Moreover, it actively reminds users of weather updates, rest reminders and hydration based on ambient conditions and personal habits.

With local deployment of ESP-Claw, UNIHIKER K10 achieves a full closed loop of local perception, proactive decision-making and hardware control. It responds instantly without cloud reliance or complex coding, turning rigid hardware into a perceptive, service-oriented intelligent desktop companion.

System Architecture of UNIHIKER K10 + ESP-Claw

In this project, UNIHIKER K10 combined with the ESP-Claw framework builds a complete embedded local AI agent system.

UNIHIKER K10 serves as the perception and interaction layer. Equipped with onboard sensors including temperature and humidity, ambient light, three-axis accelerometer, camera and microphone, as well as output modules such as display, speaker and RGB LED, it enables direct interaction between the system and the real world. It collects real-time environmental data and supports user input.

As the core brain of the system, the ESP-Claw AI agent framework runs entirely locally on UNIHIKER K10. It undertakes intent understanding, task planning, memory management, reasoning and decision-making, forming a closed loop of execution and feedback. This allows UNIHIKER K10 to independently analyze environmental status, understand user intentions and trigger corresponding actions.

With this system architecture, UNIHIKER K10 and ESP-Claw implement a full local AI agent, completing the loop of Perception → Understanding → Decision-Making → Execution. The device can not only respond instantly to environmental changes and user commands, but also proactively broadcast daily reminders and environmental information, functioning as a multimodal, locally intelligent and proactively responsive AI desktop butler.

Implementation Steps

Flashing the ESP-Claw Firmware

To run the ESP-Claw local AI framework on UNIHIKER K10, the first step is to flash the firmware onto the board.

Use a USB cable to connect your computer to UNIHIKER K10.

Visit the official ESP-Claw website at: https://esp-claw.com/en/flash/

Click Connect, select the corresponding serial port of UNIHIKER K10, and finish the connection.

After successful pairing, the following page will be displayed.

Select the chip as esp32-s3, the series as dfrobot, and the board type as dfrobot_k10.

Click the Flash Firmware button, and the website will start downloading and flashing the firmware to UNIHIKER K10.

Before downloading, please ensure your computer is connected to a stable network to obtain the latest firmware smoothly. Do not disconnect the USB cable during the flashing process to avoid firmware corruption.

Wait until the firmware download and flashing are completed.

After flashing is complete, the page will prompt that the operation succeeded.

Click Reconnect Device to reconnect UNIHIKER K10, and verify that the firmware has been properly flashed and can start up normally.

Configure Network, LLM and Communication Mode

After the firmware is successfully flashed, UNIHIKER K10 will display the following page.

Open the network settings and find the espclaw Wi‑Fi network shown on the screen of UNIHIKER K10.

Connect your phone or computer to this WiFi, wait a few seconds, and the configuration page will open automatically. You can also manually visit http://esp-claw.local/ to access the settings.

Configure Network: Click Basic Settings, enter your Wi-Fi name and password, then click Save.

After saving, unplug and replug the device to restart it, then refresh the webpage.

Once rebooted, the screen will show that Wi-Fi is connected successfully.

This indicates UNIHIKER K10 has accessed the network, and you may proceed with subsequent system configuration.

Configure LLM: Click LLM Settings, select the LLM provider and model version, then enter the corresponding API Key.

After successful configuration, the ESP-Claw framework on UNIHIKER K10 can leverage the LLM to provide natural language understanding and generation capabilities for text interaction.

Configure Communication Mode: The ESP-Claw framework supports access to various instant messaging apps, enabling UNIHIKER K10 to exchange messages across different platforms.

After finishing the LLM configuration, proceed to set up the instant messaging software. Click IM Settings, then configure the messaging platforms you need to connect. It currently supports Telegram, QQ Bot (OpenClaw), Lark, WeChat ClawBot, and more. Multiple platforms can be filled in at the same time.

Configure Search Engine: After setting up the search engine, ESP-Claw can retrieve up-to-date information via the network. Weather queries also rely on search engine support. Click Web Search Settings, select the corresponding search engine, and enter your API key.

Configure Desktop Assistant Role

After completing the configuration of the network, LLM, instant messaging software and search engine, you can set role attributes for UNIHIKER K10. This enables it to actively perceive the environment, understand user intentions, and perform hardware actions as well as information feedback.

The role prompt is the core configuration that defines the behaviors and capabilities of the desktop AI assistant Jarvis. It sets the identity, personality, abilities and interaction rules for UNIHIKER K10.

Here is the prompt I designed for the desktop AI assistant.

In this prompt, I define the interfaces of the onboard hardware on UNIHIKER K10. This allows the AI to understand commands and establishes a mapping rule from Jarvis’s language to hardware actions, enabling real-time control of devices such as the fan, RGB lighting, and screen display.

Click Memory , paste the above role prompt into the Identity field, then click Save.

By setting the desktop assistant prompt, UniHiker K10 turns into Xiaoxing, a considerate desktop butler with active response capabilities. It can not only control the fan, lights and screen, but also check the weather and set alarm reminders.

You can also chat directly with the ESP-Claw enabled UniHiker K10 via instant messaging tools such as WeChat.

You can send text messages or use voice messages to converse with the desktop assistant directly via instant messaging tools, and issue commands for it to execute.

Summary

Without any complex programming, by locally deploying the ESP-Claw framework on the UniHiker K10, we successfully equip the hardware with an AI brain capable of independent thinking, turning the device into a practical desktop AI butler.

Unlike traditional smart hardware that relies on programming and cloud-based commands to operate, this project delivers a convenient experience of direct natural language commands and instant hardware responses. Users do not need any programming skills. Simply by sending voice or text requests, they can have the UniHiker K10 perform operations such as fan control, light adjustment, and environmental monitoring. This truly upgrades the hardware from passive execution to an intelligent companion with active interaction capabilities.

Of course, leveraging the open-source nature and scalability of the ESP-Claw framework, we can design various skills to further enhance the active sensing capability of UniHiker K10 and adapt it to more diverse application scenarios.

For example, an environmental anomaly warning skill can be developed to enable automatic alerts and adjustments when temperature, humidity or light intensity exceeds preset thresholds. A user habit adaptation skill can also be created to precisely match personal schedules and deliver customized reminders for rest and work.

It is expected that more developers will further optimize and enrich this desktop AI assistant in the future.