正在加载视频...
视频加载失败
This is one of the coolest open-source AI agent projects I've seen in a while: 'Understand Anything' It's a plugin for Claude Code, Codex, OpenCode etc. that analyzes your codebase and turns it into a knowledge base that you can interact with. It explains the codebase to you, rather... show more
174,887 次观看 • 3 个月前 •via X (Twitter)
43 条评论
GitHub repo:
wow, nice. you can also try devin DeepWiki by @cognition on your repo. it produces docs based on your codebase
@DonvitoAI @cognition I've used that. Also awesome!
That is cool indeed.
Been using it all night.
The knowledge graph approach is the right one. The difference worth noting: Understand Anything is built for code and adapted for text. If your primary content is articles, PDFs, research papers, podcast notes, markdown files, you want something purpose-built for that layer. Been building exactly that. The Curator is a local open-source knowledge system that ingests any content type and builds a structured wiki automatically, entities, concepts, summaries, all interlinked. Has an MCP server so your AI agents can read and write to the knowledge graph directly during sessions. You can also visualize it as a 3D neural network in Obsidian. Not static retrieval, it understands context and the connections between ideas, not just the content itself. Free and open source, needs a Gemini or Anthropic API key for ingestion. If this sounds like the right layer for your podcast highlights workflow, would love a GitHub star, sitting at 30+ right now and every one helps with discovery.
The Obsidian test is the real signal. A graph is useful if it can answer “where did this belief come from?” with source snippets and stale nodes. For teams, that matters more than the visualization.
bro accidentally invented his own brain outside his body 888 episodes of other ppl's thoughts now haunting a graph like little knowledge ghosts 💀
拿 Obsidian 库试这个还挺对味
The next big AI unlock isn't generating code it's understanding entire systems.
this direction is more useful than most agent hype treating your own notes, research, or code as something you can converse with changes individual and small team leverage faster than trying to build fully autonomous systems. the personal knowledge management angle is underrated
a knowledge base built from your own codebase is such a smart idea. been wanting something like this for a while
For anyone sitting on years of notes, docs, or transcripts, this saves a lot of time
imagine onboarding to someone elses brain in like 30 seconds thats lowkey terrifying for the person whose vault it is 😂
I tired it but don’t understand how it works
Codebase RAG that actually preserves structure instead of flattening everything into chunks is the missing piece for agent reliability. Most current approaches lose the invariants you need for safe edits.
We just started using this tool! It's pretty incredible!
This is awesome
Yea man. Still learning new things from every podcast I've ever listened to.
Keren banget! AI agent yang bisa “understand anything” di codebase ini game changer. Bayangin kalau diintegrasikan dengan on-chain data real-time, agent bisa langsung audit smart contract + execute autonomously. Ini yang @mucriptoAI lagi eksplor. Siapa yang udah coba plugin ini?
Hopefully an end to ‘code death’ as described in the Mythical Man Month.
This is incredible
been looking for something exactly like this does it handle large codebases well or does it start to choke after a certain size?
This experiment I did was 144k lines of "code" (text in markdown). It splits it up into batches and deals with large data that way.
It’s a great utility. I ran it across my 34 repos and it’s refreshed whenever 10 files are changed in a single repo. It’s all written to an EAR on my intranet repo. It gives strategist apps a quick reference without burning tokens on codebase greps.
We’ve built this into @atomicdotdev so you can also leverage this during reviews to do a blast radius and change tracking better than git. More to come soon on a WebUI using ACP to ask, plan, modify, review, integrate changes all from a single UI with the agents you like to use
That sounds amazing! Open-source projects like this really make coding more accessible. Can't wait to see how it evolves!
Ohhh, this is neat.
Nice, thanks for sharing, man. This could be super relevant for what we're building. So yeah, super exciting.
That knowledge graph from 888 podcast episodes is such a clever use case, way beyond just code 🤯
Right!?
Right, and this works for literally any messy dataset not just code 🤯
@daniel_mac8 I need the how to get started 101
@daniel_mac8 Check the readme first, setup takes a few minutes then you just point it at any folder 🛠️
Looks like a massive shortcut. I learned the hard way it quietly kills your architectural intuition.
the gap this fills: agents get the code but never the understanding. they see files, not the why or the decisions behind them. a senior engineer's edge is the mental model they carry between sessions, and that's exactly what a persistent codebase knowledge base gives an agent.
This is super needed in the AI coding space. Reviewing diffs without being able to put together all the dots mentally in a code base is a bottleneck for serious projects.
I remember @zarazhangrui has a similar tool
does the plugin handle incremental code changes well in openCode, or do you need to rebuild the knowledge base each time? curious about scaling for active repos.
Codebase understanding is becoming infrastructure. The agent should not rediscover architecture every run; it needs a durable map of modules, ownership, constraints, and prior decisions.
@Scobleizer Oh looking nice! Need to try that one
@Scobleizer looks great and will be a real usecase for me if it works as shown
genuinely wondering how it handles drift when you add files or refactor does it re-index automatically or are you just querying stale context after a week
![Obsidian + Claude Code turned a 4,300-note vault into a second brain that answers back. Most people use Obsidian as a graveyard. You save 40 highlights a week, link nothing, and reread maybe 2% of it. The knowledge is there the retrieval is dead. Claude Code fixes the retrieval, because your vault is just a folder of markdown files and Claude Code lives in folders. The whole setup takes 20 minutes: Open a terminal inside your Obsidian folder and run Claude Code no plugins, no API glue, no export step. It reads all 4,300 files natively. Then drop a CLAUDE.md at the vault root explaining your structure: where daily notes live, how you tag, what your MOCs are. Now every session starts with context instead of chaos. From there you just talk to your vault: > "find every note where I mentioned churn and write a summary with backlinks" > "read my last 30 daily notes and list the 5 ideas I keep circling but never ship" > "build a MOC for everything tagged startup-ideas and link the orphan notes into it" It writes new notes straight into the vault with proper [[wikilinks]], and they show up in your graph view 10 seconds later. The boring layer automates too one command cleans broken links, one merges duplicates, one turns 6 months of meeting notes into a single decisions log. Work that used to eat a Sunday now runs while you make coffee. Before: 4,300 notes, 900 orphans, search by keyword and prayer. After: an agent that read everything you ever wrote and drafts in your own voice, from your own sources, with receipts. People pay $30/month for AI note apps with 10% of this. You already own the other 90%. Your notes stopped being storage. They started being staff.](https://image.24vids.com/tw-2078404662626074695/amplify_video_thumb/2078389656312217600/img/htCh1Y9vO-Wt4ERa.jpg)
