Robyn AndersonCEO, Jam Ai · AI Implementation & Business Development Leader

I hope this clears things up

27 Jul 2026 · 1 min

Watch on YouTube

1:25

In this short advice clip, Robyn Anderson explains why she prefers using Obsidian over Google Drive when building AI agents with large language models (LLMs) like Claude. She outlines three main reasons: (1) Obsidian stores local markdown files that LLMs can read directly, avoiding the extra token cost of using a Google Drive connector; (2) Obsidian provides a custom indexing system where each agent folder contains markdown instructions and per‑document blurbs, enabling fast, token‑efficient retrieval; and (3) Obsidian offers a clean, dedicated file system separate from personal or team files, reducing clutter and privacy concerns. She concludes by inviting viewers to follow her for more practical LLM and AI‑agent tips.

Chapters

I had a client ask me today if she could just use her Google Drive instead of Obsidian. And at first I said yes, but then I changed my mind. Here's why. Number one, if you're using Obsidian, these are markdown files on your computer that your large language model can read really easily. Whereas if she's using Google Drive, it's going to use the connector and it's gonna use more tokens to go in and search and find what it's looking for. Also, Obsidian creates an indexing system. That is specific for large language models. So the top of every single agent in my Obsidian folder has its own MD instructions. And so every document that's put into my agent has an index and a little blurb of what's on that document or note. So basically my large language model can go in and really easily find what it's looking for and create content really fast without using a ton of my tokens. Finally, I like having a file system that is specific for my large language model, whereas like Google Drive, I have team members in there, I have personal information, I have tax information. So while I could just share a folder inside of it, I like having a really clean, separate system for my large language model. Hopefully that helped. Give me a follow for more tips on how to use Claude and how to create AI agents that actually help you do work.