Documentation

Complete guide to using CodePathfinder for semantic code search and indexing.

Contents

1. Getting Started

CodePathfinder helps you index and semantically search your GitHub repositories. Once indexed, you can use Claude Desktop or the API to search your codebase using natural language.

What is CodePathfinder?

CodePathfinder is a semantic code search platform that:

  • Creates vector embeddings and indexes from your GitHub repositories for intelligent search
  • Uses AI-powered semantic search to find relevant code by meaning, not just keywords
  • Integrates with Claude Desktop via MCP (Model Context Protocol)
  • Provides 6 powerful tools for code exploration and analysis
  • Supports 17+ programming languages and file formats natively
  • Includes a built-in Chat interface for conversational code exploration

Quick Start Guide

  1. Join the Private Beta
    Request early access to CodePathfinder.
  2. Create a Project
    Navigate to Projects and click + New Project.
  3. Configure Your Repository
    Enter your repository URL and GitHub token (for private repos).
  4. Run the Indexer
    Click Save & Run to start indexing. Your code will be analyzed and vector embeddings created.
  5. Connect Claude Desktop
    Generate an API key and configure Claude Desktop.

2. Supported Languages

CodePathfinder provides native syntax analysis and intelligent parsing for a wide range of programming languages and file formats.

Programming Languages

🐍 Python 📘 TypeScript 🟨 JavaScript ☕ Java 🐹 Go ⚙️ C ⚙️ C++ 🗄️ SQL

Configuration & Data Formats

📋 YAML 📋 JSON 🐘 Gradle ⚙️ Properties

Documentation & Templates

📝 Markdown 🔧 Handlebars 📄 Plain Text

Note: All supported file types get intelligent chunking and vector embeddings for semantic search. Programming languages additionally receive syntax-aware analysis for functions, classes, and symbols.

3. Use Cases & Examples

Here's how teams are using CodePathfinder to supercharge their development workflows.

🚀 Developer Onboarding

New team members can ask natural language questions about unfamiliar codebases:

Example queries:

  • "How does user authentication work in this project?"
  • "Where are the API endpoints defined?"
  • "Show me how database migrations are handled"
  • "What testing framework is used and where are the tests?"

🔍 Code Discovery

Find code by describing what it does, not by guessing variable names:

Example queries:

  • "Find code that sends email notifications"
  • "Where is the payment processing logic?"
  • "Show me functions that validate user input"
  • "Find all places that interact with the cache"

📊 Code Analysis

Get insights about your codebase structure and patterns:

Example queries:

  • "List all public API functions in the user module"
  • "Find functions that don't have documentation"
  • "What classes inherit from BaseModel?"
  • "Show me all usages of the deprecated function X"

🛡️ Security & Compliance

Identify potential security concerns and compliance issues:

Example queries:

  • "Find places where user passwords are handled"
  • "Show me all SQL queries that use string concatenation"
  • "Where are API keys or secrets referenced?"
  • "Find code that handles file uploads"

📝 Documentation Generation

Identify gaps in documentation and generate content:

Example queries:

  • "Find all public functions without docstrings"
  • "Generate documentation for the UserService class"
  • "Summarize what the payment module does"
  • "Create a README section explaining the API structure"

🤖 AI Agent Integration

Give your AI agents deep understanding of your codebase:

Use cases:

  • Claude Desktop MCP integration for conversational code exploration
  • Custom AI workflows that need codebase context
  • Automated code review with full project understanding
  • Building code-aware chatbots for your development team

4. Project Management

Projects are the core of CodePathfinder. Each project represents a GitHub repository that you want to index and search.

Creating a Project

Field Description
Name A descriptive name for your project
Repository URL Full HTTPS URL to your GitHub repository
GitHub Token Personal Access Token for private repos
Branch Specific branch to index (optional)

Hint: For private repositories, you'll need a GitHub Personal Access Token with repo scope.

5. API & MCP Tool Calling

CodePathfinder provides a REST API and MCP (Model Context Protocol) integration for programmatic access to your indexed code. MCP is supported by Claude, ChatGPT, Google Gemini, and other AI agents.

🟣 Claude Desktop

Add this to your Claude Desktop config file (claude_desktop_config.json):

{
  "mcpServers": {
    "codepathfinder": {
      "command": "npx",
      "args": ["-y", "@grabowskit/mcp-bridge"],
      "env": {
        "CODEPATHFINDER_API_KEY": "YOUR_API_KEY",
        "CODEPATHFINDER_API_ENDPOINT": "https://codepathfinder.com/api/v1/mcp/tools/call/"
      }
    }
  }
}

🟢 ChatGPT (Developer Mode)

ChatGPT supports MCP through Developer Mode (available for Pro, Plus, Business, and Enterprise accounts):

  1. Go to Settings → Apps & Connectors → Advanced settings
  2. Toggle Developer mode ON
  3. In the Connectors tab, click Create app
  4. Configure with your CodePathfinder MCP endpoint: 
    URL: https://codepathfinder.com/api/v1/mcp/
Authorization: Bearer YOUR_API_KEY
  5. In conversations, select Developer mode from the "+" menu and choose your connector

🔵 Google Gemini

Gemini supports MCP through the Gemini CLI and Gemini Code Assist:

Gemini CLI Configuration

Add to your Gemini CLI config:

{
  "mcpServers": {
    "codepathfinder": {
      "command": "npx",
      "args": ["-y", "@grabowskit/mcp-bridge"],
      "env": {
        "CODEPATHFINDER_API_KEY": "YOUR_API_KEY",
        "CODEPATHFINDER_API_ENDPOINT": "https://codepathfinder.com/api/v1/mcp/tools/call/"
      }
    }
  }
}

🔧 Direct API Access

For custom integrations or other AI agents, use our REST API directly:

POST https://codepathfinder.com/api/v1/mcp/tools/call/
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "name": "semantic_code_search",
  "arguments": {
    "query": "authentication middleware"
  }
}

Available MCP Tools

  • semantic_code_search - Natural language code search
    Find code by describing what it does: "authentication middleware", "database connection handling"
  • map_symbols_by_query - Find functions, classes, methods
    Search for specific symbols: "all functions in UserService", "classes that extend BaseModel"
  • symbol_analysis - Analyze definitions and usages
    Get detailed information about a symbol: where it's defined, where it's used, its documentation
  • read_file_from_chunks - Read complete files
    Reconstruct and read any indexed file from the repository
  • document_symbols - Find undocumented code
    Identify functions and classes that lack documentation
  • size - Get project statistics
    View indexed file counts, storage size, and coverage information

6. Chat API

The Chat API enables programmatic access to the conversational code exploration interface with streaming responses (SSE).

Endpoints

  • POST /api/v1/chat/send/ - Send message, receive streaming response
  • GET /api/v1/chat/conversations/ - List conversations
  • GET /api/v1/chat/conversations/<id>/ - Get conversation details
  • DELETE /api/v1/chat/conversations/<id>/ - Delete conversation

All API requests require a project API key: Authorization: Bearer cpf_your_api_key

7. Troubleshooting

Common issues and how to resolve them.

Indexer is stuck in "Running" status

Click Stop, check the Console for errors, and try running again. Enable "Clean Index" if the problem persists.

Private repository indexing fails

Ensure your GitHub token has repo scope and hasn't expired.

Search returns no results

Verify the project is indexed (status should be "Completed"). Check that the project visibility toggle is ON.

Claude Desktop can't connect

Verify your API key is correct and hasn't been revoked. Ensure the MCP bridge config has the correct endpoint URL.

Some files aren't being indexed

Check if the file type is supported (see Supported Languages section). Very large files may be skipped. Binary files and common directories like node_modules are excluded by default.

Ready to get started?

Join the Private Beta and unlock the power of semantic code search.

Join Private Beta 🚀