How to Self-Host OpenManus AI | Requirements & Installation Guide

Since its launch, Manus AI has caused a huge stir. The first general AI agent, Manus, is unlike its competitors in many different ways. It is not just another chatbot. It goes further than other Artificial Intelligence models. It transcends the boundary of generating, assisting, or suggesting and completes the tasks given to it. 

Artificial General Intelligence (AGI) has been talked about for a while and Manus is seemingly the first step towards AGI. Manus can asynchronously and precisely do tasks assigned to it. Tasks such as screening resumes, researching, analyzing, and many more. It is a huge step towards Human-AI collaboration. Manus at the moment of writing is operating in invite-only mode; however, they have released an Open-Source version of Manus called OpenManus for everyone to use. You can deploy OpenManus locally which ensures your privacy and flexibility.  

OpenManus can be deployed with light or heavy AI models based on your requirements. OpenManus is powered by the state of the art LLMs such as Claude 3.7 Sonnet or GPT-4o. However, based on your requirements, you can use lighter, smaller LLMs and host OpenManus locally.  

Step-by-step guide on how to self-host OpenManus

OpenManus can be deployed in 2 different ways. You can either use an API from any of the big AI companies such as OpenAI, Anthropic, Deepseek, etcetera or you can use an LLM that you developed by yourself with your own requirements and preferences. For both of these methods, the first few steps will be the same. Only the configuration part will differ as you will need different codes for these methods. 

We have used uv to install the packages here. uv is a fast and efficient package installer, resolver and manager.

Step 1: uv Installation

curl -LsSf https://astral.sh/uv/install.sh | sh

Step 2: Cloning the Repository

git clone https://github.com/mannaandpoem/OpenManus.git

cd OpenManus

Step 3: Creating and Activating a New Virtual Environment

uv venv –python 3.12

source .venv/bin/activate  # On Unix/macOS

# Or on Windows:

# .venv\Scripts\activate

Step 4: Installing the Dependencies

uv pip install -r requirements.txt

Step 5: Browser Automation Tool Installation

https://github.com/mannaandpoem/OpenManus

*This step is optional as not all tasks will require browser interactions. If you want to keep your OpenManus lightweight and your tasks do not include browser interactions, then you can skip this part. 

You can also use Conda to create a virtual environment. However, we recommend using uv as it is more efficient for installing Python dependencies. 

Create a new Conda virtual environment

conda create -n open_manus python=3.12

conda activate open_manus

Cloning the Repository 

git clone https://github.com/mannaandpoem/OpenManus.git

cd OpenManus

Installing the Dependencies

pip install -r requirements.txt

Step 6: Configuration

Here’s the tricky part. First, we will show you how to host OpenManus using APIs. We have used an API from OpenAI. You can use anything else based on your preference. 

Step 1: Create a config.toml file in the config directory.

cp config/config.example.toml config/config.toml

Step 2: Add your API keys and Customized Settings by Editing config/config.toml 

# Global LLM configuration

[llm]

model = “gpt-4o”

base_url = “https://api.openai.com/v1“

api_key = “sk-…”  # Replace with your actual API key

max_tokens = 4096

temperature = 0.0

# Optional configuration for specific LLM models

[llm.vision]

model = “gpt-4o”

base_url = “https://api.openai.com/v1“

api_key = “sk-…”  # Replace with your actual API key

With this, you are done configuring your OpenManus with APIs. If you want to use your own LLM, follow this method. This method uses Ollama to run OpenManus locally. 

# Global LLM configuration

[llm] #OLLAMA:

api_type = ‘ollama’

model = “llama3.2”                                                 # The LLM model to use

base_url = “http://localhost:11434/v1”                             # API endpoint URL

api_key = “ollama”                                                 # Your API key

max_tokens = 4096                                                  # Maximum number of tokens in the response

temperature = 0.0                                                  # Controls randomness

[llm.vision] #OLLAMA VISION:

api_type = ‘ollama’

model = “llama3.2-vision”                                          # The vision model to use

base_url = “http://localhost:11434/v1”                             # API endpoint URL for vision model

api_key = “ollama”                                                 # Your API key for vision model

max_tokens = 4096                                                  # Maximum number of tokens in the response

temperature = 0.0                                                  # Controls randomness for vision model

With that, you have successfully installed OpenManus and it is ready to perform any task you require. 

To deploy your LLM using Ollama, visit How to self-host DeepSeek for more information. 

Quick Start

To quickly start OpenManus with a single line 

python main.py

And input your idea through the terminal.

Also, you can use 

python run_mcp.py

For the MCP tool version. 

And if you are using the experimental multi-agent version, you can use this

python run_flow.py

For more information, visit here. 

Hardware Requirements

The requirement for hardware will be based on the usage of your LLM. It is very easy to determine the hardware requirement with this formula: 

M = ((P x 4B)/(32/Q)) x 1.2

Here,

M = Required VRAM of GPU

P = Number of parameters in the model

4B = 4 bytes per parameter

Q = Number of bits per parameter when loading

1.2 = A 20% overhead factor

With this formula, you can easily calculate the hardware requirements to host OpenManus with your preferred LLM. 

For example, if we are using a 15b version of an LLM. Using this formula, we can calculate the hardware requirements of this LLM. 

Here,

B = 15,000,000,000

Q = 8

M = (15,000,000,000 x 4)/(32/8)) x 1.2

    = 18 GB

To run an LLM model which uses 15 billion parameters will require 18gb of GPU memory. Calculate the hardware requirements using this formula and set up your server based on that. Make sure to have at least the minimum amount of hardware. For optimum performance, build your server with better hardware than the requirements so that you are prepared if you want to upgrade. 

If you are using an API from a third-party LLM such as OpenAI or Anthropic, you will only need the minimum specs to run OpenManus. It could be anything from 2gb of GPU and upwards. Since the LLM is already hosted by the API provider, you do not need too much computational power to host OpenManus locally in this method.  

Conclusion

AI agents are becoming smarter everyday. It is important to be up-to-date with the latest AI technology. OpenManus is the first of many general AI agents. It can complete tasks without guidance from the users. Run OpenManus with proper hardware and use it to your advantage. Whether you use it for coding, business, or anything else, OpenManus can be a valuable asset to you for a long while.