wingetCollections
  • Auto
  • Search
  • Awesome Collections
  • Awesome Programs
  • My Favorite Collections
  • My Collections
  • Account / Settings
  • Terms of Service
  • Privacy
  • About

© 2026 winget.ragerworks.com

Loading...
Install Context Builder using Winget - wingetCollections
  1. Go back
  2. Packages
  3. Context Builder
Context Builder logo

Context Builder igorls

ai cli codebase context developer-tools llm rust tree-sitter
Use this command to install Context Builder:
winget install --id=igorls.context-builder -e

Context Builder is a command-line tool designed to automate the creation of a structured, LLM-optimized markdown file from any codebase directory. It streamlines the process of preparing code for AI consumption by generating clean, relevant context with minimal effort.

Key Features:

  • Relevance-Based File Ordering: Organizes files to prioritize config, project docs, source code (entry points first), tests, documentation, and lockfiles last for optimal LLM understanding.
  • AST-Aware Code Signatures: Utilizes Tree-Sitter to extract function/class signatures and structural summaries, enhancing context clarity.
  • Token Budgeting: Automatically caps output size with adjustable limits to prevent exceeding model context windows.
  • Parallel Processing: Leverages multiple CPU cores to process thousands of files efficiently.
  • Configuration Support: Allows customization via a context-builder.toml file for consistent outputs across projects.
  • Diff Tracking: Generates diffs between snapshots to highlight changes, reducing redundant content and token usage.

Audience & Benefit: Ideal for developers, data scientists, and AI engineers who need to provide context to LLMs. Context Builder saves time by automating the creation of structured, optimized documentation, ensuring AI models receive high-quality inputs tailored for accurate responses.

Available via winget for easy installation, Context Builder is a powerful tool for anyone working with codebases and large language models.

README

Context Builder

A blazing-fast CLI for creating LLM context from your entire codebase.

Crates.io Crates.io Size Deps.rs Crate Dependencies (latest) Crates.io Total Downloads

Coverage Status CI docs.rs

License

Tired of manually copy-pasting files into your LLM prompts? Context Builder automates this tedious process, creating a single, clean, and context-rich markdown file from any directory.

demo

Why Context Builder?

Providing broad context to Large Language Models (LLMs) is key to getting high-quality, relevant responses. This tool was built to solve one problem exceptionally well: packaging your project's source code into a clean, LLM-friendly format with zero fuss.

It's a command-line utility that recursively processes directories and creates comprehensive markdown documentation, optimized for AI conversations.

Core Features

  • ⚡ Blazing Fast & Parallel by Default: Processes thousands of files in seconds by leveraging all available CPU cores.

  • 🧠 Smart & Efficient File Discovery: Respects .gitignore and custom ignore patterns out-of-the-box using optimized, parallel directory traversal. Automatically excludes common heavy directories (node_modules, dist, build, __pycache__, .venv, vendor, etc.) even without a .git directory.

  • 📊 Relevance-Based File Ordering: Files appear in LLM-optimized order: config & project docs first, then source code (entry points before helpers), tests, documentation, build/CI files, and lockfiles last. This helps LLMs build a mental model faster.

Versions
0.9.0
0.8.3-1
0.8.2
Website
github.com
License
Last updated
6/5/2026
Download latest version

  • 💰 Context Budgeting (--max-tokens): Cap token output to fit your model's context window. Warns when output exceeds 128K tokens with actionable suggestions.

  • 💾 Memory-Efficient Streaming: Handles massive files with ease by reading and writing line-by-line, keeping memory usage low.

  • 🌳 Clear File Tree Visualization: Generates an easy-to-read directory structure at the top of the output file.

  • 🔍 Powerful Filtering & Preview: Easily include only the file extensions you need and use the instant --preview mode to see what will be processed.

  • ⚙️ Configuration-First: Use a context-builder.toml file to store your preferences for consistent, repeatable outputs. Initialize a new config file with --init, which will detect the major file types in your project (respecting .gitignore patterns) and suggest appropriate filters.

  • 🔁 Automatic Per-File Diffs: When enabled, automatically generates a clean, noise-reduced diff showing what changed between snapshots.

  • ✂️ Diff-Only Mode: Output only the change summary and modified file diffs—no full file bodies—to minimize token usage.

  • 🌲 Tree-Sitter AST Analysis (optional): Extract function/class signatures (--signatures), structural summaries (--structure), and smart AST-boundary truncation (--truncate smart). Supports Rust, JavaScript, TypeScript, Python, Go, Java, C, and C++.

  • 🧪 Accurate Token Counting: Get real tokenizer–based estimates with --token-count to plan your prompt budgets.

    • Installation

    Quick Install (Linux/macOS)

    Pre-built binaries include full Tree-Sitter AST support. The installer verifies SHA256 checksums before installing.

    curl -sSL https://raw.githubusercontent.com/igorls/context-builder/master/install.sh | bash

    Windows (PowerShell)

    Invoke-WebRequest -Uri "https://github.com/igorls/context-builder/releases/latest/download/context-builder-x86_64-pc-windows-msvc.zip" -OutFile "$env:TEMP\cb.zip"
Expand-Archive "$env:TEMP\cb.zip" -DestinationPath "$env:LOCALAPPDATA\Programs\context-builder" -Force
$env:PATH += ";$env:LOCALAPPDATA\Programs\context-builder"

    > Add %LOCALAPPDATA%\Programs\context-builder to your PATH permanently via System Settings.

    From crates.io

    cargo install context-builder --features tree-sitter-all

    From source

    git clone https://github.com/igorls/context-builder.git
cd context-builder
cargo install --path .

    Usage

    Basic Usage

    # Initialize a new context-builder.toml config file with automatically detected file types (respecting .gitignore)
context-builder --init

# Process current directory and create output.md
context-builder

# Process a specific directory
context-builder -d /path/to/project

# Specify an output file
context-builder -d /path/to/project -o documentation.md

    Advanced Options

    # Filter by file extensions (e.g., only Rust and TOML files)
context-builder -f rs -f toml

# Ignore specific folders/files by name
context-builder -i target -i node_modules -i .git

# Cap output to a token budget (prevents context overflow)
context-builder --max-tokens 100000

# Preview mode (shows the file tree without generating output)
context-builder --preview

# Token count mode (accurately count the total token count of the final document using a real tokenizer.)
context-builder --token-count

# Add line numbers to all code blocks
context-builder --line-numbers

# Stream the document to stdout and pipe it straight into an LLM tool
context-builder -f rs -o - | llm

# Skip all confirmation prompts (auto-answer yes)
context-builder --yes

# Output only diffs (requires auto-diff & timestamped output)
context-builder --diff-only


# Clear cached project state (resets auto-diff baseline & removes stored state)

context-builder --clear-cache

# Combine multiple options for a powerful workflow
context-builder -d ./src -f rs -f toml -i tests --line-numbers --max-tokens 100000 -o rust_context.md

    Configuration

    For more complex projects, you can use a context-builder.toml file in your project's root directory to store your preferences. This is great for ensuring consistent outputs and avoiding repetitive command-line flags.

    Example context-builder.toml

    # Default output file name
output = "context.md"

# Default output folder
output_folder = "docs/context"

# Create timestamped versions of the output file (e.g., context_20250912123000.md)
timestamped_output = true

# Automatically compute per-file diffs against the previous timestamped snapshot
auto_diff = true

# Emit only change summary + modified file diffs (omit full file bodies)
# Set to true to greatly reduce token usage when you just need what's changed.
diff_only = false

# Number of context lines to show around changes in diffs (default: 3)
diff_context_lines = 5

# File extensions to include
filter = ["rs", "toml", "md"]

# Folders or file names to ignore
ignore = ["target", "node_modules", ".git"]

# Add line numbers to code blocks
line_numbers = true

# Preview mode: only show file tree without generating output
preview = false

# Token counting mode
token_count = false

# Tokenizer encoding for --token-count / --max-tokens
# Options: "o200k_base" (GPT-4o / o-series, default) or "cl100k_base" (GPT-4 / GPT-3.5)
encoding = "o200k_base"


# Automatically answer yes to all prompts

yes = false



# Encoding handling strategy for non-UTF-8 files

# Options: "detect" (default), "strict", "skip"

encoding_strategy = "detect"

    You can initialize a new configuration file using the --init command. This will create a context-builder.toml file in your current directory with sensible defaults based on the file types detected in your project. The filter suggestions will be automatically tailored to your project's most common file extensions while respecting .gitignore patterns and common ignore directories like target, node_modules, etc. This makes it more likely to include the files you actually want to process.

    Auto-diff

    When using timestamped_output = true together with auto_diff = true, Context Builder compares the previous canonical snapshot to the newly generated one and produces:

    • A Change Summary (Added / Removed / Modified files)
    • A File Differences section containing only modified files (added & removed are summarized but not diffed)

    If you also set diff_only = true (or pass --diff-only), the full “## Files” section is omitted to conserve tokens: you get just the header + tree, the Change Summary, and per-file diffs for modified files.

    Note: Command-line arguments will always override the settings in the configuration file.

    Command Line Options

    • -d, --input - Directory path to process (default: current directory).
    • -o, --output - Output file path (default: output.md). Use - to stream the document to stdout (e.g. context-builder -o - | llm); progress messages then go to stderr so the pipe stays clean.
    • -f, --filter - File extensions to include (can be used multiple times).
    • -i, --ignore - Folder or file names to ignore (can be used multiple times).
    • --max-tokens - Maximum token budget for the output. Files are truncated/skipped when exceeded.
    • --preview - Preview mode: only show the file tree, don't generate output.
    • --token-count - Token count mode: accurately count the total token count of the final document using a real tokenizer.
    • --line-numbers - Add line numbers to code blocks in the output.
    • -y, --yes - Automatically answer yes to all prompts (skip confirmation dialogs).
    • --diff-only - With auto-diff + timestamped output, output only change summary + modified file diffs (omit full file bodies).
    • --clear-cache - Remove stored state used for auto-diff; next run becomes a fresh baseline.
    • --signatures - Replace full file content with extracted function/class signatures (requires tree-sitter).
    • --structure - Append structural summary (function/class counts) to each file (requires tree-sitter).
    • --truncate - Truncation strategy for --max-tokens: smart (AST-boundary aware, default) or byte (requires tree-sitter).
    • --visibility - Filter extracted signatures by visibility: all (default), public, or private (requires tree-sitter).
    • --encoding - Tokenizer used for --token-count and --max-tokens: o200k_base (GPT-4o / o-series, default) or cl100k_base (GPT-4 / GPT-3.5).
    • --init - Initialize a new context-builder.toml config file.
    • -h, --help - Show help information.

    Token Counting

    Context Builder uses the tiktoken-rs library to provide accurate token counts. By default it uses the o200k_base encoding, which matches GPT-4o and the o-series (and is a close approximation for current frontier models). Use --encoding cl100k_base for GPT-4 / GPT-3.5. The selected encoding applies to both --token-count and --max-tokens budgeting.

    Documentation

    • DEVELOPMENT.md: For contributors. Covers setup, testing, linting, and release process.
    • BENCHMARKS.md: For performance enthusiasts. Details on running benchmarks and generating datasets.
    • CHANGELOG.md: A complete history of releases and changes.

    Contributing

    Contributions are welcome! Please see DEVELOPMENT.md for setup instructions and guidelines. For major changes, please open an issue first to discuss what you would like to change.

    Changelog

    See CHANGELOG.md for a complete history of releases and changes.

    License

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