CortexON: Open source multi-agent AI tool to support automation of complex tasks

General Introduction

CortexON is an open-source multi-agent AI system hosted on GitHub at https://github.com/TheAgenticAI/CortexOn. It was developed by the TheAgenticAI team and has been recognized by the Manus and OpenAI DeepResearch-inspired, with the goal of automating complex tasks through the collaboration of multiple specialized agents. Users can download the code and deploy it for free for research, business processes, and technical operations.CortexON supports Docker installations, incorporates real-time search, file management, code generation, and more, and the community is constantly refining it.

Function List

• web proxy: Searching web pages in real time, acquiring data, and handling online interactions.
• file broker: Managing documents, including organizing, extracting information and storing.
• code agent: Generate code, support multiple languages, and also debug and optimize.
• executing agent: Coordinate tasks and manage multiple agent collaborations to ensure smooth processes.
• API proxy: Connect to external services and APIs to extend automation capabilities.
• multitasking: Multiple agents running concurrently to handle complex work such as research or business analysis.

Using Help

CortexON requires users to download the code and deploy it themselves. Detailed installation and usage instructions are provided below.

Installation process

1. Download Code
   Open a terminal and enter the command Clone Repository:

git clone https://github.com/TheAgenticAI/CortexOn.git

Go to the project catalog:

cd CortexOn

2. Configuring Environment Variables
   In the project root directory, create the .env file, fill in the following required variables:

• Anthropic API(for AI modeling):

  ANTHROPIC_MODEL_NAME=claude-3-7-sonnet-20250219
  ANTHROPIC_API_KEY=你的Anthropic密钥
  

  Get the key from https://console.anthropic.com/.

• Browserbase Configuration(for web proxies):

  BROWSERBASE_API_KEY=你的Browserbase密钥
  BROWSERBASE_PROJECT_ID=你的项目ID
  

  Register and create a project at https://browserbase.com/.

• Google Search(for the search function):

  GOOGLE_API_KEY=你的Google密钥
  GOOGLE_CX=你的搜索ID
  

  Configured at https://developers.google.com/custom-search/v1/overview.

• Logging(Optional):

  LOGFIRE_TOKEN=你的Logfire密钥
  

  Retrieved from https://pydantic.dev/logfire

• WebSocket(Front-end communication):

  VITE_WEBSOCKET_URL=ws://localhost:8081/ws
  

3. Installing Docker
   Make sure that Docker Desktop is installed on your computer. if not, download and install it from https://www.docker.com/products/docker-desktop/.
4. Enable host network (optional)
   In Docker Desktop settings, open "Network" under "Resources", check "Enable host networking "and restart Docker.
5. Build and run
   Enter the following command in the terminal:

docker-compose build
docker-compose up

When started, the service runs on the local port. Access the following address:

• Front-end interface:http://localhost:3000
• Backend API:http://localhost:8081(Documentation in http://localhost:8081/docs)
• Agentic Browser:http://localhost:8000(Documentation in http://localhost:8000/docs)

6. Verify Installation
   Open your browser and visit http://localhost:3000If you see the interface, it means the installation is successful.

Main Functions

Web Proxy: Search the Web

• move::

1. Open the front-end interface (http://localhost:3000).
2. Enter a keyword in the search box, such as "artificial intelligence trends".
3. Click "Search" and the Web Agent will return the results.

• take note of: Internet connection required. Results depend on keyword accuracy.

Documentation agent: managing documents

• move::

1. Select "File Management" in the interface.
2. Upload files (supports PDF, text, etc.).
3. Set rules such as "Sort by date" or "Extract summary".
4. Click "Execute" and the file will be automatically organized.

• exports: The results are saved by default in the project directory's output Folder.

Code Agent: Generate Code

• move::

1. Go to the "Code Generation" module.
2. Enter a requirement such as "Write a Java sorting program".
3. Select language (Python, Java, etc. are supported).
4. Click "Generate" to view the code.
5. If there are problems, click "Debug" to optimize.

• save (a file etc) (computing): You can download the code or copy it for use.

API proxy: connecting to external services

• move::

1. Select "API Integration" in the interface.
2. Enter the external API address and parameters (e.g. weather API).
3. Click "Connect" and the agent will get the data.

• use: Can be used for extended functionality, such as access to real-time data.

Executive agent: multitasking collaboration

• move::

1. Create tasks in Task Management, such as "Analyze market data".
2. Setting up subtasks: web agent search, file agent organization, code agent generation of charts.
3. Click "Run" and the agent will collaborate to complete.

• tempo: The interface shows the status of the task, and the results are automatically saved upon completion.

Featured Function Operation

Dynamic multi-agent collaboration

At the heart of CortexON are multiple agents working together. For example, doing a research paper:

1. Web Agent Search Profile.
2. Document agent organizes content.
3. The code agent generates visual charts.
4. The executive agent is integrated into the report.
   Input the target in the interface, click "Start", the system will finish automatically.

Open Source and Technology Stack

• customizable: The user can modify the code. Open the project folder, adjust it with the editor and re-run it docker-compose upThe
• technology stack::
• Framework: PydanticAI
• Browser: Browserbase
• Search: Google SERP
• Backend: FastAPI
• Front-end: React/TypeScript, TailwindCSS
• Log: Pydantic Logfire

application scenario

1. academic research
   When you need to write a paper, CortexON saves time by searching the literature, organizing your notes, and generating diagrams.
2. business automation
   Companies use it to collect market data, generate reports, and streamline processes.
3. Developer Tools
   Programmers use code agents to generate code snippets, or debug programs, to improve efficiency.

QA

1. Does CortexON charge a fee?
   No charge. It is an open source project and the code is free. But you need to provide your own server and API key.
2. Need to know how to program?
   Installation requires knowledge of basic commands, using the interface does not.
3. Do you support Chinese tasks?
   Support. The interface and documentation are in English, but can handle Chinese search and file tasks.