What's holding me back from AI repos and agents isn't running it locally though. Its the lack of granular control. I'm not even sure what I want. I certainly don't want to approve every request, but the idea of large amounts of personal data being accessible, unchecked, to an AI is concerning.
I think perhaps an agent that focuses just on security, that learns about your personal preferences, is what might be needed.
Agreed regarding the privacy/security hesitations. Running the models locally with ollama is an option, but of course there's the hardware requirements and limitations of open source models to contend with. ultimately it's a balance between privacy and ease of use, and I'm not sure that there's a good one-size-fits-all for that balance.
is your idea of granular control (roughly) a group of agents in separate containers writing back to their own designated store each sufficient, or more control than that?
Great work. As someone who spends many hours a day in Claude Code and dreads the dreaded auto compact moment, the memory problem is genuinely a big point of frustration.
Right now I use a skill on every commit (or when the auto compact warning starts showing up) that forces Claude to update its "memory" . It is a flat markdown file that gets stuffed into conversations, not v smart. Claude forgets things I've told it dozens of times.
Your MCP server approach makes total sense. The create_atom tool alongside semantic_search makes it read/write from day one. I would love to wire a stop hook to automatically atomize session insights (the write side). That's the dream: I work on something in my code, Claude learns why, and that knowledge flows into Atomic without me saying "remember this."
Thanks! Integrating Atomic with tools like claude code is one of the more exciting use cases in my opinion. There are a lot of tools for AI memory out there, but not a ton that allow you to browse, organize, and collaborate directly with the memories.
The honest answer is it doesn't help a ton, at least not in its current form. It's fun to look at, and occasionally I'll see some interesting semantic connections between articles - but by far the more useful tools here are the wiki generation, auto-tagging, and chat/MCP features. The graph view definitely needs more love - if anyone has thoughts on how to make it more useful, I'd love to hear them.
I like the headless approach here. Since you already have hierarchical auto-tagging, do those categories act as "gravitational anchors" for the spatial canvas to prevent a "semantic hairball" once the knowledge base scales beyond a few hundred atoms?
Not 100% sure what are the ingestions methods available ?
Browser extension clipper and RSS are two. I guess I can manually create a node/atom ? Can it scan a local folder for markdown notes ? Or ocr some pdf -> markdown/frontmatter sidecar files -> atomic node ? That would be the dream.
Yes, regarding ingestion methods the following are available:
- RSS feed
- Web clipper browser extension (working on publishing the extension to the chrome store)
- Import a folder of markdown files (desktop app only)
- Manually ingest a one-off URL
- Manually create an atom
The iOS app is a more recent addition, also not yet published to the app store, but the idea is to add a share target so that during mobile browsing you can quickly add articles to your KB.
PDF support would be great, and I'd love to hear more ideas about ingestion methods if anyone has them!
Yeah, the data layer is sqlite - not just the vectors but the notes themselves, wiki and chat data, etc. The project started out as solely a desktop app, so sqlite seemed a natural fit. Since then I've pivoted to a server/client model, but the purpose is still very much for personal use so I still feel like sqlite fits that niche.
Thanks! The project is still early stages, haven't had a chance to get the app signing set up - right now the easiest way to get started is using the web interface via docker compose.
For sure. The idea here (or at least how I've been using it) is to use Atomic as catchall place to put personal notes, interesting articles, research ideas .. pretty much anything, and Atomic will handle the categorization and knowledge synthesis. For example, I have a knowledge base that uses RSS to sync top Hacker News articles and I'll occasionally generate new wiki-style articles which summarize and synthesize the articles based on top-level categories (AI, hardware, philosophy, you name it).
I think tools like this will get really popular once more non-technical users get comfortable with CLI-based agentic tools. What's your go-to agent harness when using this? Will check it out!
Clean approach to connecting knowledge semantically. The self-hosted angle is smart — data ownership matters especially for personal knowledge. How are you handling the semantic matching under the hood?
This is pretty cool!
What's holding me back from AI repos and agents isn't running it locally though. Its the lack of granular control. I'm not even sure what I want. I certainly don't want to approve every request, but the idea of large amounts of personal data being accessible, unchecked, to an AI is concerning.
I think perhaps an agent that focuses just on security, that learns about your personal preferences, is what might be needed.
Thanks for taking a look!
Agreed regarding the privacy/security hesitations. Running the models locally with ollama is an option, but of course there's the hardware requirements and limitations of open source models to contend with. ultimately it's a balance between privacy and ease of use, and I'm not sure that there's a good one-size-fits-all for that balance.
is your idea of granular control (roughly) a group of agents in separate containers writing back to their own designated store each sufficient, or more control than that?
Is this something like what you want?
greywall.io
Great work. As someone who spends many hours a day in Claude Code and dreads the dreaded auto compact moment, the memory problem is genuinely a big point of frustration.
Right now I use a skill on every commit (or when the auto compact warning starts showing up) that forces Claude to update its "memory" . It is a flat markdown file that gets stuffed into conversations, not v smart. Claude forgets things I've told it dozens of times.
Your MCP server approach makes total sense. The create_atom tool alongside semantic_search makes it read/write from day one. I would love to wire a stop hook to automatically atomize session insights (the write side). That's the dream: I work on something in my code, Claude learns why, and that knowledge flows into Atomic without me saying "remember this."
Thanks! Integrating Atomic with tools like claude code is one of the more exciting use cases in my opinion. There are a lot of tools for AI memory out there, but not a ton that allow you to browse, organize, and collaborate directly with the memories.
Why not implement this on top of obsidian?
Does anybody mind explaining how the web of articles in the first image helps the writer?
The honest answer is it doesn't help a ton, at least not in its current form. It's fun to look at, and occasionally I'll see some interesting semantic connections between articles - but by far the more useful tools here are the wiki generation, auto-tagging, and chat/MCP features. The graph view definitely needs more love - if anyone has thoughts on how to make it more useful, I'd love to hear them.
Thank you for clarifying. Also, great work! I am looking forward to trying it out for my personal learning note taking this week.
My first thought is that this would be a great place to find new topics of interest at the "overlap" regions.
I like the headless approach here. Since you already have hierarchical auto-tagging, do those categories act as "gravitational anchors" for the spatial canvas to prevent a "semantic hairball" once the knowledge base scales beyond a few hundred atoms?
Nice ! Congratulations.
Not 100% sure what are the ingestions methods available ? Browser extension clipper and RSS are two. I guess I can manually create a node/atom ? Can it scan a local folder for markdown notes ? Or ocr some pdf -> markdown/frontmatter sidecar files -> atomic node ? That would be the dream.
Thanks, much appreciated!
Yes, regarding ingestion methods the following are available:
- RSS feed
- Web clipper browser extension (working on publishing the extension to the chrome store)
- Import a folder of markdown files (desktop app only)
- Manually ingest a one-off URL
- Manually create an atom
The iOS app is a more recent addition, also not yet published to the app store, but the idea is to add a share target so that during mobile browsing you can quickly add articles to your KB.
PDF support would be great, and I'd love to hear more ideas about ingestion methods if anyone has them!
This looks pretty neat! Thanks for sharing ^_^
I saw sqlite-vec for semantic search so I assume notes are stored in sqlite.
- What considerations did you have for the storage layer?
- Also does storage on disk increase linearly as notes/atoms grow?
Thanks!
Yeah, the data layer is sqlite - not just the vectors but the notes themselves, wiki and chat data, etc. The project started out as solely a desktop app, so sqlite seemed a natural fit. Since then I've pivoted to a server/client model, but the purpose is still very much for personal use so I still feel like sqlite fits that niche.
[dead]
[dead]
I did something similar, markdown and code agents for memory, multiple feeds for intake, also my own browsing and claude cli messages get indexed.
Great work, but your macOS build cannot be opened. You need to sign the app through Apple Developer Program.
System Settings > Privacy & Security, scroll down to the "Security" section, and click Open Anyway
Thanks! The project is still early stages, haven't had a chance to get the app signing set up - right now the easiest way to get started is using the web interface via docker compose.
Just following up here, I set up app signing so the newest Atomic version (v1.1.0) should no longer have this issue.
Seems like a LogSeq/Roam/Obsidian alternative?
For sure. The idea here (or at least how I've been using it) is to use Atomic as catchall place to put personal notes, interesting articles, research ideas .. pretty much anything, and Atomic will handle the categorization and knowledge synthesis. For example, I have a knowledge base that uses RSS to sync top Hacker News articles and I'll occasionally generate new wiki-style articles which summarize and synthesize the articles based on top-level categories (AI, hardware, philosophy, you name it).
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
looks fine
I think tools like this will get really popular once more non-technical users get comfortable with CLI-based agentic tools. What's your go-to agent harness when using this? Will check it out!
Clean approach to connecting knowledge semantically. The self-hosted angle is smart — data ownership matters especially for personal knowledge. How are you handling the semantic matching under the hood?