I dug up a replacement for Deep Research and Manus, which is a newly released open source project.

There is Deep Research, which is only available to Pro members, and Manus, which can only be experienced with a sky-high invitation code. AI agent is indeed a very eye-catching application field.

For people who often deal with papers and materials, a powerful AI assistant can indeed save a lot of time and energy.

Recently, I discovered a new project on GitHub - CortexON, which says its mission is to be a fully open source alternative to Deep Research and Manus.

What is CortexON?

CortexON integrates multiple specialized agents at the bottom layer, which dynamically collaborate to achieve user-defined goals. These specialized agents include:

• Web Agent : Handles real-time Internet search, data retrieval, and network interaction.
• File Agent : Manages file operations, organization, data extraction, and storage tasks.
• Coder Agent : Generates, debugs, and optimizes code snippets in various programming languages.
• Executor Agent : Executes tasks, manages workflow, and coordinates communication between agents.
• API Agent : Seamlessly integrate with external services, APIs, and third-party software to expand automation capabilities.

These agents combine their unique capabilities through dynamic coordination to efficiently automate complex tasks.

 How it works

Key Capabilities

• Advanced, context-aware research automation
• Dynamic multi-agent orchestration
• Seamless integration with third-party APIs and services
• Code Generation, Debugging, and Execution
• Efficient file and data management
• Personalized and interactive task execution, such as travel planning, market analysis, educational content creation, and business intelligence

Technology Stack

CortexON is built using the following technologies:

• Framework : PydanticAI Multi-Agent Framework
• Headless Browser : Browserbase (Web Proxy)
• Search Engine : Google SERP
• Logging and Observability : Pydantic Logfire
• Backend : FastAPI
• Front-end : React/TypeScript, TailwindCSS, Shadcn

How to use

Environment variables

Create one .env file and add the following required variables:

Anthropic API

• ANTHROPIC_MODEL_NAME=claude-3-7-sonnet-20250219
• ANTHROPIC_API_KEY=your_anthropic_api_key
• Get an API key from Anthropic (https://console.anthropic.com/).

Browserbase Configuration

• BROWSERBASE_API_KEY=your_browserbase_api_key
• BROWSERBASE_PROJECT_ID=your_browserbase_project_id
• Set up an account and project on Browsebase (https://www.browserbase.com/).

Google Custom Search

• GOOGLE_API_KEY=your_google_api_key
• GOOGLE_CX=your_google_cx_id
• Follow the steps on the Google Custom Search API (https://developers.google.com/custom-search/v1/introduction).

Logging

• LOGFIRE_TOKEN=your_logfire_token
• Create a token on LogFire (https://logfire.com/).

WebSocket

• VITE_WEBSOCKET_URL=ws://localhost:8081/ws

Docker Setup

Follow these steps to set up and run the project using Docker:

1. Clone the CortexON repository :

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

• Make sure you have added the .env The files are placed in CortexOn in the directory.

• Enable host networking in the Docker Desktop settings. See the Docker Desktop Networking guide (https://docs.docker.com/desktop/networking/#use-cases-and-workarounds) for more information.

docker-compose build
docker-compose up

• Build the image and start the container.

Access Services

Once the container is running, you can access the service through the following URL:

• Frontend : http://localhost:3000
• CortexON backend : http://localhost:8081
• API documentation: http://localhost:8081/docs
• Agentic browser : http://localhost:8000
• API documentation: http://localhost:8000/docs

Notice :

• Need to your_anthropic_api_key,your_browserbase_api_key Replace the placeholders with your actual API key and ID.
  Make sure it is running docker-compose up Before,.env All variables in the file are set correctly.
  If you are using Docker Desktop, follow the guide to enable host networking to ensure that the network is configured correctly.