TokenScope

TokenScope

A Model Context Protocol server that enables token-aware directory exploration and file analysis for LLMs, helping them understand codebases through intelligent scanning and reporting.

Category
Visit Server

README

TokenScope

TokenScope is a token-aware directory explorer for Large Language Models.

A Model Context Protocol (MCP) server for token-aware directory exploration and analysis, designed for Large Language Models (LLMs).

Overview

TokenScope provides intelligent directory structure analysis and token-aware file content exploration. It helps LLMs like Claude understand codebases and directory structures by:

  1. Scanning directory structures with token-efficient summaries
  2. Extracting and analyzing file contents with token awareness
  3. Finding important files for codebase understanding
  4. Generating reports with relevant information

Features

  • Token-Aware Directory Scanning

    • Explores directories recursively with configurable depth
    • Provides intelligent summaries for large directories
    • Respects .gitignore files and custom ignore patterns
  • File Content Analysis

    • Smart extraction of file contents that respects token limits
    • Special handling for JSON and other structured files
    • File selection prioritization based on importance
  • Token Usage Statistics

    • Estimates tokens required to process directories
    • Breaks down token usage by file extension
    • Identifies token-heavy files
  • Comprehensive Reporting

    • Generates markdown reports with directory structure
    • Includes token usage statistics
    • Shows samples of important files
  • Security Features

    • Path validation to restrict operations to a specified base directory
    • Prevents access to files outside the allowed base path

Installation

Prerequisites

  • Python 3.10 or higher
  • uv (recommended for easy dependency management)

1. Main Installation (PyPI)

This is the recommended method for most users who just want to use TokenScope:

# Install from PyPI using uv (recommended)
uv pip install tokenscope

Running TokenScope

The --base-path argument is mandatory for security reasons. It restricts all file operations to the specified directory.

# Run using the installed package
uv run --with tokenscope tokenscope --base-path /path/to/allowed/directory

Configuring in Claude Desktop

  1. Locate Claude Desktop's configuration file (typically in ~/.config/claude/config.json)

  2. Add TokenScope to the mcpServers section:

    "mcpServers": {
      "TokenScope": {
        "command": "uv",
        "args": [
          "run",
          "--with",
          "tokenscope",
          "tokenscope",
          "--base-path",
          "/your/secure/base/path"
        ]
      }
    }
    
  3. Replace /your/secure/base/path with the directory you want to restrict operations to

  4. Save the configuration file and restart Claude Desktop

2. Development Installation (from GitHub)

For contributors or users who want to modify the code:

# Clone the repository
git clone https://github.com/cdgaete/token-scope-mcp.git
cd token-scope-mcp

# Install development dependencies with uv
uv pip install -e ".[dev]"

Running in Development Mode

# Run the server directly with uv
uv run --with fastmcp --with tiktoken src/server.py --base-path /path/to/allowed/directory

Configuring Development Version in Claude Desktop

  1. Locate Claude Desktop's configuration file

  2. Add TokenScope to the mcpServers section with development paths:

    "mcpServers": {
      "TokenScope (Dev)": {
        "command": "uv",
        "args": [
          "run",
          "--with",
          "fastmcp",
          "--with",
          "tiktoken",
          "/path/to/your/token-scope-mcp/src/server.py",
          "--base-path",
          "/your/secure/base/path"
        ]
      }
    }
    
  3. Replace /path/to/your/token-scope-mcp/src/server.py with the actual path to the server.py file

  4. Replace /your/secure/base/path with your secure directory

Security Features

The --base-path argument is mandatory for security reasons:

  • All file operations are validated to ensure they're within the specified directory
  • Attempts to access or modify files outside the base path will be rejected
  • The base path is set once when starting the server and cannot be changed without restart

Example Prompts

Here are some examples of how to use TokenScope with Claude:

Please scan my project directory at /path/to/project and tell me about its structure, focusing on the most important files.
Analyze the token usage in my project directory at /path/to/project and tell me how many tokens would be needed to process the entire codebase with an LLM.
Generate a comprehensive directory report about my project at /path/to/project, including structure, token statistics, and samples of the most important files.

Available Tools

The server provides the following MCP tools:

scan_directory_structure

Scans a directory and returns its structure in a token-efficient way.

scan_directory_structure(
    path: str, 
    depth: int = 3,
    max_tokens: int = 10000,
    ignore_patterns: list[str] | None = None,
    include_gitignore: bool = True,
    include_default_ignores: bool = True
)

extract_file_content

Extracts the content of a specific file, respecting token limits and format.

extract_file_content(
    file_path: str, 
    max_tokens: int = 10000,
    sample_only: bool = False
)

search_files_by_pattern

Searches for files matching specified patterns within a directory structure.

search_files_by_pattern(
    directory: str,
    patterns: list[str],
    max_depth: int = 5,
    include_content: bool = False,
    max_files: int = 100,
    max_tokens_per_file: int = 1000,
    sample_only: bool = False,
    ignore_patterns: list[str] | None = None,
    include_gitignore: bool = True,
    include_default_ignores: bool = True
)

analyze_token_usage

Analyzes token usage for a directory or file to estimate LLM processing requirements.

analyze_token_usage(
    path: str,
    include_file_details: bool = False,
    ignore_patterns: list[str] | None = None,
    include_gitignore: bool = True,
    include_default_ignores: bool = True
)

generate_directory_report

Generates a comprehensive markdown report about a directory with token statistics.

generate_directory_report(
    directory: str, 
    depth: int = 3,
    include_file_content: bool = True,
    max_files_with_content: int = 5,
    max_tokens_per_file: int = 1000,
    sample_only: bool = False,
    ignore_patterns: list[str] | None = None,
    include_gitignore: bool = True,
    include_default_ignores: bool = True
)

copy_file_to_destination

Copy a file from source path to destination path.

copy_file_to_destination(
    source_path: str,
    destination_path: str
)

Default Ignore Patterns

TokenScope automatically ignores common directories and files:

DEFAULT_IGNORE_PATTERNS = [
    ".git/",
    ".venv/",
    "venv/",
    "__pycache__/",
    "node_modules/",
    ".pytest_cache/",
    ".ipynb_checkpoints/",
    ".DS_Store",
    "*.pyc",
    "*.pyo",
    "*.pyd",
    "*.so",
    "*.dll",
    "*.class",
    "build/",
    "dist/",
    "*.egg-info/",
    ".tox/",
    ".coverage",
    ".idea/",
    ".vscode/",
    ".mypy_cache/",
]

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

  • Built with FastMCP
  • Uses tiktoken for accurate token counting
  • This same concept was implemented originally in repoai
  • Inspired by the need to efficiently analyze codebases with LLMs

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured