Which AI feature would you want first in Memos?

[bluedbird]

We’ve been thinking about AI features for Memos, but want to keep it fast, lightweight, and focused.

If AI were added, what would actually help you capture or revisit thoughts better?

Examples:

• speech-to-text for faster capture
• auto tags
• related memos suggestions
• short summaries (per day or tag)
• ...

We’re especially interested in ideas that improve quick capture, without turning Memos into a heavy AI workspace.

Share your use case if possible — that helps us prioritize.

[coffandro]

Is there a reason to add LLM's into memos? Like if there's a reason behind it sure but it seems to just be adding LLM for LLM's sake, no?

[subcidr]

Jumping on this too, we really do not need AI in Memos. There is enough use if AI in all sorts of apps these days, getting tired of it. Memos is lean and nice as it is, I'd much rather want to see focus on polish and features that are already on the roadmap.
Please no AI.

[boojack]

Fully agree with NO AI — but that should mean no AI chat, no AI-generated text content, etc.

For my personal use case, I need a faster way to capture thoughts through voice input, ideally with speech to text, so that it can replace typing in some scenarios. Not sure if this aligns.

[nrdwr-timmmi]

I’d really like it if the software offered a really good way of managing things via MCP. Then I could set my AI agents to work on memos whenever I want.

tl;dr: It’s better to implement a good MCP interface

[brashica]

Fully agree with NO AI — but that should mean no AI chat, no AI-generated text content, etc.

For my personal use case, I need a faster way to capture thoughts through voice input, ideally with speech to text, so that it can replace typing in some scenarios. Not sure if this aligns.

An (optional) integrated Whisper mode or something similar for dictation seems like a great idea. I'm using a Whisper model on FUTO keyboard for this purpose and its been very nice to have. It would be great to have that ability incorporated so it works on any platform through the browser.

[brashica]

Auto note tagging or importing convos with an llm and that kind of thing gives me concern because they are easily not mindful. You can quickly end up with a lot of slop a human didn't write or care to read afterwards. There are also privacy concerns if the model is not running locally. Maybe to some this doesn't matter but personally I would not want to hand over the content of my notes to a big tech company.

[lincolnthalles]

• Speech-to-text can become a killer feature.

• Auto tags are probably the simplest, yet still useful feature.

Related memo suggestions and summaries can be wasteful in many instances. It would need caching and extra knobs to keep token usage under control.

[Mysteriza]

AI-powered auto-tags are a great idea—I really like that. And I’d also love it if AI could be leveraged further, for example, to improve the structure of the notes we write so they’re neater, clearer, and conform to the required Markdown structure. Of course, this would be a feature that can be turned on or off—it’s optional—and users would need to click a “button” to initiate the refinement process and set up their own AI API key to make it work.
Even better, use Ollama for local AI.

[ncguk]

Speaking for myself only, I would prefer no A.I. features. In my opinion A.I. is unreliable, a privacy problem, and I wouldn't want it anywhere near my private data. If you do decide to add A.I. features, please also add the ability to completely disable them.

[flyingcakes85]

+1
*completely disable at build time

[GamerPET]

Personally I don't want AI powered tags because I don't want Memos to become a dump of things. I have Karakeep for that and Karakeep is using AI Tags, they just become dump tags.

One thing that I did was createa n8n workflow that captures my "Past Month" memos, summarize each memos per tag, then create a MD file with them. Basically I have a #nas where I write all the things I did on my NAS. This will take all my NAS notes and summarize them. And he does that for each tag, monthly.

Memos -> Ollama model -> MD File -> Post into Memos that I have a new summary -> ntfy push to my phone (my current flow)

[musaabhasan]

If AI is added to Memos, I would keep it opt-in, local-first where possible, and focused on capture/retrieval rather than turning the app into a general AI workspace.

The safest first features are probably:

• speech-to-text for fast capture
• suggested tags that the user can accept, edit, or ignore
• related-memo suggestions that run only on selected notes or trusted spaces
• weekly/daily summaries generated on demand, not continuously in the background

For privacy, the important controls are: per-workspace AI enable/disable, per-memo exclusion, provider selection, clear indication when content leaves the device, and user-editable derived artifacts. Auto-tags and summaries should be stored as suggestions or metadata with provenance, not silently mixed into the original note.

A useful research/knowledge-work flow would be bilingual or multilingual note discovery: users capture rough notes in one language, then retrieve related memos across languages without needing to manually normalize every tag. That is practical, but still narrow enough to fit the lightweight nature of Memos.

[mdg-pnw]

Instead of adding AI features, can we get multi-workspace support?
https://github.com/orgs/usememos/discussions/4155

I'd love to be able to share this with groups of my friends, but it seems like sharing is an all or nothing approach at the moment, and it prevents me from actually sharing this with people.

[ZangXincz]

Notes fully organized and logged by AI tend to turn our notebooks into disorganized clutter. Content churned out without our own critical reflection rarely delivers meaningful value.
As for AI-powered features, I believe AI search would be an extremely practical addition. At present, we can only search via exact keywords. Once our note library grows to a certain scale, two common pain points emerge: either we can’t recall the exact keyword for the information we need, or we input a keyword yet fail to pull up relevant entries.
I’m hoping semantic AI search can be implemented, which retrieves memos based on the semantic meaning of our input keywords or full questions.
Keyword semantic matching: If we search the keyword "weight loss", the system will surface all memos with contextually relevant content, rather than only exact matches.
Question-based retrieval: When we pose a full question like "How to lose weight quickly and safely", the tool will locate all memos containing related ideas.
This feature would be incredibly useful. We take notes not merely for the sake of recording things down—more importantly, to support our future selves. And the core of that support lies in being able to instantly pull up the right information whenever we need it.

[icouckuy]

Little bit late here, but I was looking for an alternative to my current note taking app (joplin; there are many many others). I just need something to jug down an idea, interesting items (to buy), new artists to listen to, etc. Memos seemed a better fit and nice for that (I guess a don't forget the milk app works too, but slightly different focus). While I can add proper tags for searching later, all music recommendations would still be very fragmented (not one list).

Thinking about my workflow, AI/LLMs makes sense to consolidate (give an overview) of all my inner thoughts (mostly different lists of things to do, buy, research, watch, etc.). I agree it doesnt have to be integrated, but an easy way to do it (via MCP) can be nice.

[yylqing]

Actually, I don't want AI to directly create or modify my notes, but it can periodically summarize and review them, while also enhancing search content.
Right now, I'm using Hermes to receive content sent via WeChat and saving it to Memos through MCP, while recognizing attachments like images and saving them to comments; scheduled tasks periodically review and summarize this and send it to me.