packx

Bundle code context enforces a strict 49k token limit by default, filtering and compressing your codebase into a single output file. Use `packx --limit 49k -c src/` to create a shareable, AI-ready context bundle that maximizes token efficiency for LLM prompts, code review, or debugging sessions.

Yes, packx supports comment stripping and minification to compress code before sending to LLMs. These filters reduce token overhead while preserving functionality, letting you fit more relevant code within the 49k limit for context-aware search or AI analysis.

Code bundling outputs in XML, Markdown, JSONL, or plain text formats. Each format suits different AI workflows—XML for structured parsing, Markdown for readability, JSONL for streaming ingestion, and plain text for direct prompt inclusion within your token budget.

Packx uses versatile filters and follow-imports to selectively include files matching your criteria. This lets you exclude irrelevant code, focus on specific modules, and automatically trace dependencies—all while respecting the 49k token limit.

A strict token limit prevents prompt injection risks, ensures predictable LLM response quality, and keeps API costs controlled. The 49k default in packx balances comprehensive context with safety and efficiency for reliable AI-assisted code tasks.

No, packx's follow-imports feature automatically traces and includes dependencies in your bundle. This ensures your AI context captures the full import graph while staying within the 49k limit, eliminating manual dependency tracking.