Skip to main content

How to Use AutoGLM for Free: A Complete Open-Source Mobile AI Agent Guide

BlogsAutoGLMMobile AI AgentAI AutomationAndroid AutomationOpen Source AIMobile AgentAI Phone ControlAI Workflow AutomationLLM AgentsFree AI ToolsAbout 4 min

How to Use AutoGLM for Free: A Complete Open-Source Mobile AI Agent Guide

Zhipu AI quietly released something remarkable: AutoGLM, their fully open-source mobile AI Agent framework. With this release, anyone can now build an AI system that can see a phone screen, understand what’s happening, and control the device using natural language instructions—completely for free.

In other words, you can type a simple command like:

“Open YouTube and search for AI tutorials.”

And your Android phone will actually perform the task by itself.

This guide walks you through how to use AutoGLM from scratch, with no paid API keys and no cloud subscription required. Everything runs locally under your control.

Official GitHub Repository:
https://github.com/zai-org/Open-AutoGLM

1. What Is AutoGLM?

AutoGLM is an open-source mobile AI Agent system built on top of multimodal large language models. Instead of relying on fragile automation scripts, it allows an AI to:

  • Visually understand what is displayed on the phone screen
  • Decide which action should be taken next
  • Automatically tap, swipe, and type
  • Execute full multi-step tasks autonomously

In simple terms, AutoGLM turns your smartphone into a truly AI-operated device, not just a chatbot interface.

Since it supports local deployment, you also get:

  • No recurring API costs
  • No forced cloud dependency
  • No automatic data upload

You remain fully in control of your device and your AI system.

2. What You Need to Use AutoGLM for Free

To get started, you only need:

  • A Windows, macOS, or Linux computer
  • An Android phone running Android 7.0 or later
  • A USB cable or Wi-Fi connection for ADB
  • Python 3.10 or newer
  • Basic command-line experience

A dedicated GPU is recommended for smooth model performance, but CPU-only setups can still work for testing and learning.

3. Step 1: Install Python

If Python is not installed yet, download it from:

https://www.python.org

After installation, verify it:

python --version

Step 2: Install ADB (Android Debug Bridge)

ADB allows your computer to issue commands directly to your phone.

Download it from Google’s official website:

https://developer.android.com/tools/releases/platform-tools

After installation, verify:

adb version

5. Step 3: Enable Developer Mode on Your Phone

On your Android device:

  1. Open Settings

  2. Go to About Phone

  3. Tap Build Number multiple times until Developer Mode is enabled

  4. Enable USB Debugging

Connect your phone and verify:

adb devices

If your device appears, the connection is successful.

6. Step 4: Install ADB Keyboard (For Automatic Typing)

AutoGLM requires a special input method to simulate typing.

  • Download ADB Keyboard APK from the AutoGLM repository

  • Install it on your phone

  • Activate it in Input Method Settings

This enables the AI to automatically type inside any application.

7. Step 5: Install AutoGLM

Clone the repository:

git clone https://github.com/zai-org/Open-AutoGLM.git
cd Open-AutoGLM

Install dependencies:

pip install -r requirements.txt
pip install -e .

8. Step 6: Download and Run the AutoGLM Model

AutoGLM currently offers two official models:

  • AutoGLM-Phone-9B (optimized for Chinese)

  • AutoGLM-Phone-9B-Multilingual (recommended for global users)

You can deploy the model locally using vLLM:

python3 -m vllm.entrypoints.openai.api_server \
 --served-model-name autoglm-phone-9b \
 --model zai-org/AutoGLM-Phone-9B \
 --port 8000

Once running, your local AI API will be available at:

http://localhost:8000/v1

At this point, you already have a fully functional local AI engine for mobile control.

If you plan to run AutoGLM continuously (for example, 24/7 mobile automation or remote device control), hosting your model service on a lightweight cloud VPS can be very convenient. Many developers use hourly-billed VPS platforms such as LightNode for this purpose, since you can scale up or shut down instantly without long-term cost commitments.

9. Step 7: Control Your Phone with Natural Language

Interactive Mode

python main.py --base-url http://localhost:8000/v1 --model "autoglm-phone-9b"

Then type a command such as:

Open Chrome and search for OpenAI

Your phone will perform the action automatically.

One-Line Command Mode

python main.py --base-url http://localhost:8000/v1 "Open TikTok and scroll videos"

Python API Mode

from phone_agent import PhoneAgent
from phone_agent.model import ModelConfig

model_config = ModelConfig(
    base_url="http://localhost:8000/v1",
    model_name="autoglm-phone-9b",
)

agent = PhoneAgent(model_config=model_config)
agent.run("Open Amazon and search for wireless headphones")

This mode is ideal for developers who want to integrate AutoGLM into larger automation systems.

10. Wireless Phone Control Without USB

You can also control your phone wirelessly:

adb connect 192.168.1.88:5555

Once connected, you can run AutoGLM without any physical cable. This setup is commonly used for:

  • Remote device testing

  • Cloud-based phone automation

  • Multi-device mobile AI farms

In real-world deployments, many users place both the model service and control scripts on cloud servers. For short-term testing environments, VPS services that support hourly billing (such as LightNode) are often chosen to avoid resource waste.

11. Supported Applications

AutoGLM already supports dozens of mainstream applications, including:

  • Messaging: WhatsApp, WeChat, Telegram

  • E-commerce: Amazon, eBay, AliExpress

  • Navigation & Services: Google Maps, food delivery apps

  • Media: YouTube, TikTok, streaming platforms

View the full supported list:

python main.py --list-apps

12. Real-World Use Cases

AutoGLM is suitable for many practical scenarios, such as:

  • Automated content browsing

  • Product price monitoring

  • App workflow testing

  • Accessibility support for elderly and disabled users

  • Mobile AI research

  • Large-scale phone automation

It effectively converts smartphones from manual tools into AI-operable devices.

Frequently Asked Questions (FAQ)

Is AutoGLM completely free to use?
Yes. AutoGLM is fully open-source and does not require any paid APIs or subscriptions for local deployment.

Do I need a GPU to run AutoGLM?
A GPU is recommended for better performance, but CPU-only setups can still work for basic testing.

Does AutoGLM support iPhone?
No. AutoGLM relies on ADB, which only works with Android devices.

Is my phone data uploaded to the cloud?
No. All recognition and reasoning happen locally unless you manually connect to an external API.

Can AutoGLM be used in commercial projects?
Yes, commercial usage is allowed under the open-source license. However, automated control of third-party apps must comply with platform rules and local regulations.

Find Cheap and Best servers.