FolioDux cuts token usage by 94% with file-mapping standard

FolioDux v1.0 was released on 2026-06-18, consisting of a markdown file (FOLIODUX.md), a Node-JS CLI, and prompt templates for Claude, ChatGPT, Gemini, Cursor, and Copilot [DevTo][GitHub]. The markdown spec defines three sections—TASKS, INDEX, and GROUPS—each mapping concise keywords to concrete source files. The CLI scans a project's package.json, classifies every file by type, extracts top-level comments, and writes a ready-to-commit FOLIODUX.md in seconds [GitHub].

By reducing the average token count per request from ~5,000 to ~300, FolioDux cuts OpenAI or Anthropic API bills by an order of magnitude for typical CRUD projects [DevTo]. The rule-based index guarantees that even a 128 k-token model never exceeds its window, eliminating the need for truncation or manual chunking. Because the spec is pure markdown and the CLI has zero external dependencies, any future LLM that accepts a system prompt can plug in the same index without code changes, fostering a shared convention for AI-assisted development [GitHub].

The shift from an "all-or-nothing" prompt to a deterministic, token-first contract lets small teams reap the same speed and cost benefits that large enterprises achieve with custom tooling. This contract gives the community a concrete reference point for future AI-code-navigation standards [DevTo].