MashuPack

Product Introduction

1. Definition: MashuPack is a client-side, browser-based application that functions as a structured code repository compiler and context assembler for AI-assisted development. It operates entirely within the user's browser using the File System Access API and WebAssembly.
2. Core Value Proposition: It exists to eliminate file-count limits, upload friction, and messy context assembly in browser-based AI workflows. MashuPack transforms a local codebase into a single, clean, and navigable text file optimized for consumption by large language models (LLMs) like ChatGPT, Claude, and Gemini, making code context portable and intentional.

Main Features

1. Local Browser-Based Processing: All file scanning, tree building, and processing occurs locally in the user's browser. MashuPack uses the File System Access API to obtain read-only access to a selected folder, ensuring no code or file data is ever uploaded to an external server. This guarantees maximum privacy and security for proprietary codebases.
2. Intelligent File Selection and Tree Management: The application provides a virtualized, interactive file tree. Users can expand/collapse folders, use a search bar (/ to focus), and employ shift/alt-click for bulk operations. Crucially, users can tick checkboxes for individual files or entire folders to define a precise subset of the repository for export, enabling focused context delivery to AI tools.
3. Structured Plain-Text Export with AI Navigation Markers: The core output is a single .txt file. It begins with a plain-text directory map, followed by each selected file's content wrapped in explicit START OF FILE and END OF FILE markers that include the full file path. This structure allows AI models to grep, search, and navigate the export as a virtual project, reading specific files on-demand rather than processing an undifferentiated blob.
4. Real-Time Analytics and Token Estimation: The right panel displays live project statistics: file count, folder count, total size, and root folder name. A key feature is the ability to toggle the "Size" stat between bytes and an estimated token count (using a standard ~4 characters per token heuristic). This allows developers to pre-validate that their export will fit within their target AI model's context window before generation.
5. Syntax-Highlighting File Viewer and Multi-Format Export: Clicking any file in the tree opens a dedicated File Viewer tab with syntax highlighting (powered by CodeMirror) for preview without affecting the selection. Export options include the main combined text file, a full .zip archive of the original project, direct copy-to-clipboard functionality, and a save-as-.txt option from the report panel.

Problems Solved

1. Pain Point: Inefficient and limited code context provisioning for AI coding assistants. Manually copying and pasting multiple files into an AI chat is tedious, subject to file-count limits, and loses crucial project structure, which degrades the AI's ability to understand code relationships and dependencies.
2. Target Audience: Software developers and engineers using browser-based AI tools (e.g., ChatGPT Plus, Claude.ai, Gemini Advanced) for code analysis, refactoring, debugging, or feature development. It is particularly valuable for full-stack developers, open-source contributors, and tech leads working with large or complex repositories.
3. Use Cases: Preparing a clean, structured code snapshot for an LLM to debug a complex issue; providing a focused subset of a monorepo (e.g., only frontend src/ files) for feature implementation; creating a portable project overview for technical documentation or onboarding; ensuring a code review context stays within an AI model's token limit by selectively exporting relevant modules.

Unique Advantages

1. Differentiation: Unlike simple archiving tools or naive copy-paste, MashuPack is specifically engineered for AI consumption. Its output format with explicit path-boundary markers is a deliberate innovation that enables AI navigation. Unlike cloud-based code snippet tools, it offers complete privacy through local processing. Compared to CLI tools, it requires no installation and offers a visual, interactive selection interface.
2. Key Innovation: The combination of client-side WebAssembly (Rust) for high-performance indexing of massive repositories (tested on 28,000+ file repos) and the structured plain-text export protocol with AI-friendly markers. This technical stack allows for near-instant interaction post-scan and produces an output that transforms how LLMs interact with bundled code, moving from passive ingestion to active navigation.

Frequently Asked Questions (FAQ)

1. Is MashuPack safe to use with private code? Yes. MashuPack is a static web application that runs 100% in your browser. It uses the File System Access API to read files locally. No code, file contents, or metadata is sent to any server; all processing happens on your machine. Your privacy is maintained.
2. How does MashuPack help AI models understand code better? MashuPack exports code with clear START/END OF FILE markers and full path headers. This structure allows AI models with file navigation capabilities (like Claude and ChatGPT) to search for specific file paths, read individual files, trace imports, and navigate the codebase logically, much like a developer using grep or find in a terminal, leading to more accurate and context-aware responses.
3. What is the token estimator and how accurate is it? The token estimator toggles the project size display from bytes to an estimated token count using a standard approximation of 4 characters per token. This aligns with typical tokenization for English code in models like GPT and Claude. It is a reliable heuristic for gauging context window usage but is not a precise tokenizer.
4. Can MashuPack handle very large repositories? Yes. Its core indexing engine is built in Rust and compiled to WebAssembly, running in a Web Worker to avoid blocking the browser UI. It has been tested on massive codebases like Kubernetes (~28,000 files). The initial filesystem scan time depends on your browser and hardware, but subsequent operations like selection and export remain fast.
5. What are the main export formats from MashuPack? The primary export is a single, structured plain-text file (.txt) designed for AI tools. MashuPack can also generate a standard .zip archive of the entire original project folder, independent of your selection. You can also copy the combined text directly to your clipboard for pasting into an AI chat interface instantly.