The .multilo/ folder

Every project has a hidden .multilo/folder where Multilo keeps per-project configuration and caches. It’s inspired by .claude/ and similar conventions: plain files you can read, edit, and version with your project.

.multilo/
├── AGENTS.md        # per-project instructions for the AI
├── voice.md         # your author voice
├── integrity.md     # research-integrity guardrails
├── memory/          # agent memory (one markdown file per fact)
├── index/           # on-device semantic search index (git-ignored)
├── skills/          # project skills
└── corpus/          # corpus-level research knowledge

AGENTS.md

Project-specific instructions for the AI — conventions, terminology, and rules — loaded into chat and every agent so they follow your project’s guidelines without being told each time.

voice.md & integrity.md

Your author voice and integrity rules. voice.md makes output sound like you; integrity.md enforces the academic non-negotiables.

memory/

Agent memory — one small Markdown file per remembered fact, so you can read and curate exactly what the AI knows about the project.

index/

The on-device semantic search index. It’s git-ignored (it rebuilds from your sources) but travels with the project folder.

skills/ & corpus/

skills/ holds project-specific skills the AI can use; corpus/ holds corpus-level knowledge distilled from your library for deeper, project-wide reasoning.