AI Interaction with (all) Notes
One of the best features of Mem.ai is the ability to ask the AI questions and have it answer using your library of notes. As a PhD student with thousands of notes, this feature would be indispensable and absolutely worth the money to me. It would be great if you could chat with Capacities AI and receive responses using the content of your notes across your Capacities library. For example, if you could ask the AI to consolidate and create bullet point highlights of all of your notes about a particular subject.
23 Comments
Sign in to comment
Is this a way to list all notes in a given space? I may have ideas that are in different notes because of the architecture of the system—for example, I often have different "daily notes" in a given day, on different subjects—one of these may be a true "Daily Note" but as I can only have one of these per day, the others would be simple "Pages" (sorry if I'm doing this all wrong, as I've just started with the program), or "Atomic Notes". I don't think there's any way of looking at all of the notes in a given space, as the mechanics may mean that they land in different objects although they may have relations.
Any update on when we will get full CRUD in MCP?
Like some others before me, it is becoming a bigger and bigger need to have an MCP feature set that supports not only reading all your notes but also creating and updating notes and specific object types as well. Tob build any productivity workflows and leverage AI you really need those, including properties and values would really be a huge improvement as well.
The internal AI improvements have been great, now we just need to tie it all together! :)
Thanks!
AI should absolutely be able to create tasks from (atomic) notes /content collected throughout the day. that would make it a great assistant

We've made a lot of progress on our AI roadmap. 🎉 As of recent releases, the AI assistant can now access your full space in several ways:
- Search across your space: When you ask the AI a question, it automatically searches your notes to find relevant content — even if you haven't manually selected it as context. It uses both full-text search (keyword matching) and semantic search (meaning-based matching), so a question about "dog" can also surface notes about mammals, pets, or domestic animals.
- Read note content: Once the AI finds relevant notes, it can open and read their full content, including following backlinks to related notes.
- Explore connections: The AI can traverse your knowledge graph — backlinks, property relations, and text mentions — to understand how your notes relate to each other.
- Cite sources: Every AI response includes clickable links to the notes it referenced, so you can verify and explore further.
The AI already has access to your entire space when answering questions. It searches, reads, and connects your notes to give you grounded answers with sources.
We will expand and optimize the behavior over time, but a first major milestone is achieved! 🚀
MCP Integration would be very helpful, I have been trying a simplae MCP Connector it sort of works, but has issues. I really would like to be able to pick and choose my AI tools to work with my collected information. BTW awesome product :)
- the AI should default have read all the notes. (may be by indexing)
- few skill must be included. like search, edit, create content, organize, moving , remove content (by user double confirm the action)
- the AI should have web search skill, it would be much better to search, type , organize at the same place. (we cannot ask them to search real time web data at this moment)
MCP integration would be game changing in the way to approach and manage knowledge via Capacities unique object relation model

We just released a more powerful AI Assistant. 🚀
It's not full space AI yet, but you can now have multiple notes as context, the AI can search your notes and provide links to relevant content in its answers. See the full update here: https://capacities.io/whats-new/release-59
I would go as far as to say that the AI-interaction is not good enough as long as this interaction with ALL content does not exist.
Yes we need this. The reason to keep me on Notion is for this.
Add my raised hand for this feature. It's a must have
We need this asap. As the space grows, it would be much faster, efficient and accurate if the AI could browse the entire space to provide answers.
This is a wonderful idea
May you can develop your MCP server with this task? It could give freedom for users to use any model to interact with Capacities. For some people it will be more useful than just AI in capacities. Thank you!
Excelente propuesta
A função mais esperada para começar a pagar o plano! Onde a AI faz uma leitura de todo banco de objetos no Capacities
Could we get an update on when work will start on this? I keep seeing update after update with new "alias" functionality and I don't really care about that. I'd like to see incremental movement on this instead. This is the biggest bummer I've had since moving from Obsidian. In Obsidian, I could point my Claude Code at it and use it like a document repo. I know its not that simple with Capacities but I'm sure it would make us all feel better if we could see that y'all are moving towards that goal this year.
Absolutely top feature request
This is probably my top feature request.
best
I still house some duplicated work notes on mem.ai for this reason. If something similar is done here, I'd only need capacities. For example, at work, I ask it to give me a list of projects I worked on this week and I get it so, if we could do something where we interact with our notes that would be amazing.

I find the current AI search capability very limited. It is EXTREMELY slow...and does not find the relevant information. I asked today for "what can i do to help a team member who is struggling" (as I know I have a page directly related to that). The page titled "Helping Struggling Team Members"... was NOT found by current semantic search ? How is that possible ... keyword search should have found it easily as well.
As more info: I am BYOK (tested using various Gemini models. All, including Flash were almost unbearably slow
Anyway, just want to share my experience.