README.md

Onboarding Context Generation

We follow the LLMSTxt formatting standard and offer two flexible tools to generate onboarding contexts:

Data Sources

• GeneratedOnBoardings: If your repository already contains pre-generated onboarding data, our tools will automatically detect and use it.
• Website Demo Fallback: If no generated data is available in the repo, you can create it on-the-fly via our interactive demo.

Available Tools

1. get_onboarding_context_with_code
   Produces a comprehensive context including code references. Ideal for large repositories (contexts >100K tokens).

2. get_onboarding_context_without_code
   Produces a streamlined context excluding code references. Useful when you need a lean summary.

Tool Arguments

Argument	Description
repo_name	Repository identifier (must match an entry in GeneratedOnBoardings).
token_budget	Maximum allowed tokens in the returned context. Contexts exceeding this limit will be truncated.

Quick Start

1. Using the Python Script

python utils/read_repo.py --repo_name <YOUR_REPO> --token_budget <MAX_TOKENS>

2. Running via Claude Desktop (Local MCP)

1. Install Claude Desktop if you haven’t already.

2. Configure your MCP server by editing ~/Library/Application Support/Claude/claude_desktop_config.json:

   {
     "mcpServers": {
       "codeboarding_mcp": {
         "command": "uv",
         "args": [
           "--directory",
           "/Users/<YOUR_USER>/Documents/github_folder/CodeBoarding-MCP",
           "run",
           "main.py"
         ]
       }
     }
   }

3. Launch the MCP:

   fastmcp run main.py

TODO: ngrok and setup config for remote use.