Talk to your project — chat scoped to any folder, library, or notes

Multilo's chat used to see your whole open document and your active library. Now you can scope a chat to any folder, any file, or any collection of files in the project — your library, your results/ folder, your notes.md, a single PDF — and the AI only answers from what's inside that scope. Same chat surface, much sharper grounding.

What it ships

• Talk to your library — open a chat scoped to your project library and ask "what does my library say about X?" The answer cites only your indexed references, never the open web.
• Talk to a folder — right-click any folder in the Explorer pane (your results/, data/, chapters/, anywhere) and start a scoped chat. The AI sees every file inside it, recursively.
• Talk to your notes — point a chat at a single notes.md or meeting-notes/ folder and pull out themes, action items, or open questions without re-reading the whole thing yourself.
• Talk to one document — scope a chat to a single PDF, .docx, or markdown file. Useful for long supervisor returns where you want to ask "what specifically did they ask me to change?"
• Source-anchored answers — every reply quotes the actual passages the AI used, with a click-through to the source line. No more wondering whether the model paraphrased a paper that was never in your library.

How it works

1. Pick the scope — right-click a folder or file in the Explorer, or click the folder icon in the chat header to scope an existing thread.
2. Ask anything — type a question. Multilo semantically retrieves the relevant passages from inside the scope and answers from those.
3. Click through any citation — every quoted passage links back to the exact line in the source file. Verify the AI before you trust it.
4. Mix scopes per turn — within a single chat you can re-scope between turns. Ask the library first, then narrow to a specific paper.

Why it matters

Unscoped chat tools sometimes invent. ChatGPT and Claude in their default modes have nothing to ground a claim against, so they fill gaps with plausible-sounding training data. Multilo's folder-scoped chat trades that breadth for honesty: the AI can only quote what you've actually put in front of it. If your library doesn't cover the question, the chat says so and stops — instead of fabricating a citation that doesn't exist.

Use it for

• Asking your library a question before you write the next paragraph
• Mining a notes/ folder for themes across a semester of meetings
• Pulling action items out of a long supervisor feedback document
• Summarising what's actually in a results/ folder you haven't opened in weeks
• Asking a single dense PDF the question you'd ask its author

Try it

Open any project, right-click a folder in the Explorer pane, and pick Talk to this folder. Or click the scope chip at the top of any chat to narrow it.