Your AI agent needs a map, not a memo
DOX is a zero-dependency convention that keeps AI agents oriented by turning AGENTS.md into a self-documenting hierarchy of local rules.

What it does
DOX gives AI agents spatial awareness by nesting AGENTS.md files throughout a codebase. The root file holds project-wide context, while child files define local rules for specific directories. Before editing, the agent walks the tree from root to target; after changing code, it updates the docs it just used. It is essentially a convention encoded in Markdown—no runtime, no package, no ceremony.
The interesting bit
The clever part is the feedback loop: the agent both reads and writes these files, so documentation drifts with the code rather than rotting beside it. Because DOX is just a Markdown template, it works anywhere AGENTS.md is already supported—Codex, Claude Code, OpenCode—without waiting for ecosystem buy-in.
Key highlights
- Zero runtime: a single Markdown file that agents read as instructions
- Hierarchical navigation: the agent walks from root to target, gathering context at each level
- Self-maintaining: agents update the relevant AGENTS.md files after meaningful edits
- Universal compatibility: works with any agent that already consumes AGENTS.md
- No lock-in: remove the files and the framework vanishes without a trace
Verdict
Worth trying if you already use Codex, Claude Code, or similar AGENTS.md-aware agents; everyone else will find it is simply a folder of Markdown files with no runtime to enforce anything.
Frequently asked
- What is agent0ai/dox?
- DOX is a zero-dependency convention that keeps AI agents oriented by turning AGENTS.md into a self-documenting hierarchy of local rules.
- Is dox open source?
- Yes — agent0ai/dox is open source, released under the MIT license.
- How popular is dox?
- agent0ai/dox has 1.3k stars on GitHub and is currently cooling off.
- Where can I find dox?
- agent0ai/dox is on GitHub at https://github.com/agent0ai/dox.