contextly 1.0.7

The ultimate context-engineering workspace CLI for LLMs.

Context-Ly CLI

Context-Ly CLI is an open-source Context Intelligence Engine designed to help developers generate high-quality, token-efficient context for Large Language Models (LLMs).

Rather than manually explaining your project to an AI assistant in every session, Context-Ly analyzes your repository, discovers conventions, learns team rules, and generates structured context files that help AI tools understand your codebase more effectively.

The CLI acts as a persistent Context Memory Layer for your repository, enabling consistent AI interactions across development workflows.

---

Features

• Repository analysis and context generation
• Automatic framework and dependency detection
• Architecture visualization through project structure analysis
• Team convention discovery and memory management
• Persistent project-specific context storage
• LLM-ready Context Pack generation
• Repository complexity and token usage inspection
• Context-as-Code workflow with version-controlled project memory

---

Installation

Install Context-Ly directly from PyPI:

pip install contextly

Verify the installation:

contextly --help

---

Prerequisites

• Python 3.9 or later
• A local Git repository or project directory to analyze

No external services or API keys are required for the core functionality.

---

Quick Start

Initialize Context-Ly in your project:

contextly init

Analyze your repository and generate a complete project context:

contextly analyze

This command automatically:

• Reads your README documentation
• Scans the project structure
• Detects frameworks and dependencies
• Discovers conventions and stored memory
• Generates a comprehensive PROJECT_CONTEXT.md

The generated file can be used directly with AI coding assistants and LLMs.

---

Commands

contextly init

Initialize Context-Ly in the current project.

contextly init

Creates:

.contextly/
├── config.yaml
├── memory/
└── packs/

---

contextly analyze

Generate a complete repository context file.

contextly analyze

This command:

• Reads project documentation
• Analyzes repository structure
• Detects frameworks and technologies
• Loads stored team conventions
• Generates PROJECT_CONTEXT.md

Output:

PROJECT_CONTEXT.md

---

contextly discover

Run the Pattern Discovery Engine.

contextly discover

Discovers repository conventions such as:

• TailwindCSS usage
• Zustand state management
• React Query patterns
• Service-layer architecture hints
• Framework-specific conventions

The command provides insight into patterns already present within the codebase.

---

contextly learn --auto

Convert discovered conventions into permanent project memory.

contextly learn --auto

Example:

Save convention: TailwindCSS (Uses TailwindCSS for styling.)? [y/N]

Approved conventions are stored in:

.contextly/memory/rules.yaml

This creates a persistent memory layer that can be committed to source control and shared across teams.

---

contextly memory

Inspect all stored project memory and conventions.

contextly memory

Displays all saved rules, conventions, and architectural preferences currently remembered by Context-Ly.

---

contextly pack <directory>

Generate an LLM-ready Context Pack from a specific directory.

contextly pack src/components

The command:

• Reads all files in the target directory
• Calculates token usage
• Bundles the content into a reusable Context Pack

Output location:

.contextly/packs/

Useful for sharing focused portions of a large codebase with an LLM.

---

contextly inspect

Analyze repository complexity and token consumption.

contextly inspect

Provides visibility into:

• Large files
• Potential token-heavy directories
• Context window bottlenecks
• Repository complexity hotspots

This helps identify areas that may negatively impact AI context quality.

---

contextly export <pack_name>

Fuses your memory rules and the specified context pack into a single, comprehensive Context Payload.

contextly export cli

The output is instantly copied to your clipboard, ready to be pasted directly into an LLM.

---

contextly explain <domain>

Extracts a highly-optimized structural context payload for a specific domain based on the AST Knowledge Graph.

contextly explain core

It copies a JSON payload to your clipboard, allowing the LLM to understand the architecture without wasting tokens scanning raw files.

---

Understanding Context-Ly Ignore Philosophies

Context-Ly utilizes two distinct "ignore" policies depending on the operation:

1. Packing & Inspection (contextly pack, contextly inspect): These commands respect your .gitignore and .contextlyignore files. This ensures that generated packs and token counts omit irrelevant files (like node_modules, compiled binaries, etc.), producing concise, token-efficient context for the LLM.
2. Discovery & Intelligence (contextly discover, contextly learn): The Pattern Discovery Engine ignores your .gitignore. It uses a minimal, hardcoded skip-list (only completely toxic directories like .git or .venv). This allows Context-Ly to correctly discover architectural patterns and package dependencies in valid directories (like a frontend/ folder) that you might have legitimately added to .gitignore to keep your root repository clean.

---

Example Workflow

contextly init

contextly discover

contextly learn --auto

contextly analyze

Result:

.contextly/
PROJECT_CONTEXT.md

Your repository now has a persistent memory layer and an AI-ready context file generated from both repository analysis and learned team conventions.

---

Why Context-Ly?

Modern AI coding tools are powerful, but they often lack project-specific context.

Context-Ly bridges that gap by transforming repository knowledge, team conventions, and architectural patterns into structured context that can be consistently shared with LLMs.

The goal is simple:

Build context once. Use it everywhere.

---

Changelog

For all release notes and version history, please see the CHANGELOG.md.

---

Contributing

Contributions, issues, and feature requests are welcome.

If you discover a bug, have an idea for improving repository intelligence, or want to contribute new scanners and analysis capabilities, please open an issue or submit a pull request.

---

License

This project is open source and distributed under the terms of its license.

See the LICENSE file for details.