Nexus is an autonomous AI coding agent that integrates with the Model Context Protocol (MCP) to dynamically load and manage MCP servers for extended capabilities. It operates in multiple modes (CODE, ARCHITECT, ASK) to support different workflows: CODE for direct code manipulation and execution, ARCHITECT for planning and architectural tasks with restricted access, and ASK for research and information gathering. Nexus uses LangGraph and LangChain to coordinate tool usage, plan tasks, and maintain a persistent context across sessions. When connected to MCP servers, Nexus can discover available tooling, inject tool descriptions into prompts, and orchestrate multi-server capabilities to accelerate development tasks. To start, ensure the required Python environment is active, and the Nexus MCP server is configured to load its MCP servers from the .nexus/mcp_config.json file. Nexus will then automatically load tools from connected MCP servers and make them available in its toolbox for execution within the active task context.

Prerequisites:

• Python 3.11+ (recommended 3.11–3.12 for compatibility with LangChain/LangGraph)
• Git
• Access to OpenAI or other LLM providers as configured

Installation steps:

1. Clone the repository: git clone https://github.com/DataRohit/Nexus.git cd Nexus

2. Create and activate a Python virtual environment:

   • On Unix or macOS: python3 -m venv venv source venv/bin/activate
   • On Windows: python -m venv venv .\venv\Scripts\activate

3. Install dependencies: pip install -r requirements.txt

   If a specific MCP-related dependency is listed, install it as well

   e.g., pip install mcp-python

4. Configure environment variables and MCP config:

   • Copy or create the MCP config at .nexus/mcp_config.json as needed (see mcp_config example below).
   • Set your OpenAI API key or other LLM provider credentials in the environment or config file.

5. Run the Nexus server: python -m nexus

   Or if using a specific entrypoint/module, adjust accordingly

6. Verify the server is running and MCP servers are loaded. Access the CLI or UI as documented in the repository to interact with Nexus.

Tips and common issues:

• Ensure Python virtual environment is active before starting Nexus to avoid dependency conflicts.
• The MCP configuration file (.nexus/mcp_config.json) must exist and be valid JSON to allow dynamic server loading. Use the MCP schema described in the repository docs.
• If MCP servers fail to load, check environment variables (OPENAI_API_KEY, DATABASE_URL) and ensure network access to external services.
• If using multiple MCP servers, you can define several entries under mcp_config.mcpServers and Nexus will manage them concurrently.
• For production deployments, consider persisting metrics in a dedicated database and securing API keys with a secrets manager.