ReadMe

costly_ivory_meggi

Turn your files into a memory filesystem for AI agents.

1likes0comments

Visit project 1

Project details

About this project

It doesn't make sense to ask your agent about your codebase if it cannot read the code; it must have the context.

Similarly, an agent can't give you good responses if it doesn't know about you.

ReadMe turns your files into a "codebase," so that next time you ask something, it can look up the relevant context about you before answering.

It is super easy to set up!

Just run:

codex exec --dangerously-bypass-approvals-and-sandbox --ephemeral - < Init.md

Or simply paste the content in Init.md as a prompt!

What will happen is...

Agent scans local files:

Source traces:
    ├── ~/.codex/sessions/
    ├── ~/.claude/sessions/
    ├── ...
    ├── ~/Downloads/
    └── ~/Projects/

Builds ~/.codex/user_context/

~/.codex/user_context/
└── 2026_agentic_rag_memory_systems/
    ├── year.md
    └── 2026-Q2_prompt_programs_context_retrieval/
        ├── quarter.md
        └── 2026-04_bootstrap_and_consolidation/
            ├── month.md
            └── 2026-04-13_hierarchical_memory_rewrite.md

Installs a hint in ~/.codex/AGENTS.md:

<user_context>
Additional user-specific context may live in `~/.codex/user_context/`...
</user_context>

Example question

What have I been working on in the past 3 months?

The agent reads the archive:

ls ~/.codex/user_context/2026_.../

Example answer

You have been converging on agentic infrastructure: memory systems, context retrieval, and public-facing artifacts.