frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

I reverse-engineered the three biggest agent-memory tools

2•pauliusztin•2h ago
I spent weeks reading about how Cognee, Graphiti, and Neo4j's `agent-memory` build their agent memory architectures. They converged on the same heavy knowledge-graph design: an ontology, LLM extraction pipelines, deduplication, the works.

I really wanted to use them for my personal use case, but that looks like such a heavy setup that adds a lot of friction and silos. Plus, it feels like I just get my data trapped in their service, for not a ton of value.

That's why my "long-term memory" still lives in Obsidian, Readwise, and Google Drive, with per-project LLM wikis as the agent's memory. No infrastructure. And I'm fine with it.

They ship memory as a product, which, in my opinion, at a personal or small scale, is overkill. You can build the same "knowledge graph" experience via plain old `.md` files within an LLM wiki memory.

But still, graphs are strong, so I adapted the same architecture from the Cognee, Graphiti, and Neo4j `agent-memory` stacks to build a data-mining tool with just MongoDB, VoyageAI, and Gemini Flash. But I scoped it to a very particular problem and ontology domain to avoid the KG noise.

On the other end of the spectrum, if you want to ship a product at medium-to-large scale, it makes sense to start using monsters such as Neo4j, Zep, or HydraDB.

But I am curious: what is your long-term memory setup? Obsidian + LLM wikis vs. Cognee/Graphiti/Zep? Do you actually use tools such as Cognee or Zep?

In case you are curious about how Cognee, Graphiti, and Neo4j's `agent-memory` work under the hood, I wrote a full breakdown here: https://www.decodingai.com/p/unified-memory-from-scratch-knowledge-graphs

Comments

coder-pm•58m ago
Did any of these actually covers the invalidation? It's easy to store memory and access it but I'm curious how these tools handles the fact that something is not true anymore?

GitHub is starting to lockdown some of the previously public data

5•punkpeye•28m ago•3 comments

Ask HN: Does language complexity affect reasoning?

4•zecice•49m ago•0 comments

Ask HN: What Are You Working On? (July 2026)

274•david927•2d ago•1065 comments

I reverse-engineered the three biggest agent-memory tools

2•pauliusztin•2h ago•1 comments

ChatGPT Subagents 2

2•217•2h ago•0 comments

Ask HN: Question regarding personal knowledge management

2•diarrheaasmr•4h ago•1 comments

Ask HN: Add flag for AI-generated articles

1079•levkk•2d ago•453 comments

Does disposable or temp accounts hurt the business of SaaS or app? How?

3•YPANDAR•1h ago•2 comments

Ask HN: How do you do marketing in the age of slop?

8•probst•4h ago•10 comments

Ask HN: Anyone Else Burnt Out?

5•lmpdev•6h ago•1 comments

Ask HN: Should I do a CS masters at Cambridge or start as a new grad at Amazon?

27•sspehr•1d ago•91 comments

ChatGPT Remote Connection Failure

3•tumidpandora•13h ago•2 comments

Ask HN: How useful is Gemini for Google Apps?

3•cobbzilla•14h ago•3 comments

Ask HN: US Equivalent of Anabin?

2•xqb64•1d ago•2 comments

Ask HN: How can I publish a paid app to Google Play without giving up my privacy

3•SamCoding•17h ago•4 comments

Ask HN: A DC Powered Home?

4•dhosek•18h ago•7 comments

LinkedIn is a cesspool of scammers and identity theft

79•lbpdev•9h ago•49 comments

Ask HN: How do you use Vim in the era of AI?

48•rstagi•5d ago•95 comments

Ask HN: Advice for first-year CS students?

8•jassembk•20h ago•17 comments

I want to build an OTT app. Where should I start?

4•ibuld•1d ago•3 comments

Ask HN: Any objective research on which languages are best for AI agents?

5•mwigdahl•1d ago•6 comments

Ask HN: How do you keep up with news in a particular sector?

2•doruk101•1d ago•12 comments

Ask HN: Using X vs. LinkedIn?

4•lalithaar•1d ago•4 comments

Ask HN: Can someone build open source Claude tag please?

2•lalithaar•1d ago•1 comments

Ask HN: Should I Promote My SaaS to get first 100 Customers without budget

8•HSK11•1d ago•17 comments

Tell HN: Up/downvotes of comments on HN are silently ignored when using Tor

7•fsflover•1d ago•11 comments

You've reached the end!