Links, notes, screenshots, PDFs, and videos—all the things you don't want to lose. Savedly keeps them organized and helps you find them again, even when you don't remember what you called them.
Free to use · Open source · Self-hostable
Saving things got easy. Finding them again didn't.
Every app has a share button. Every browser has a bookmark. So we save everything—articles, videos, screenshots, notes, PDFs, ideas.
Then, eventually, we need something we saved. We remember what it was about. Maybe we even remember why we saved it. But we don't remember where it went, what we called it, or which app we put it in.
The thing you're looking for isn't gone. It's just buried.
That's not really a storage problem. It's a finding problem. And folders and tags only help if you remember to organize everything yourself.
Savedly takes care of that part.
Save something once and Savedly understands what's inside. Later, find it the way you remember it—or just ask. And when it answers, it shows you exactly which saved item it came from.
Paste a link, upload a file, or send a screenshot. Savedly reads it, understands what matters, and makes it easy to find later.
You paste a link. That's the only step that's yours.
Quick Capture
fieldnotes.dev/post/rag-chunks
The page is fetched and its text pulled out. Images are read, PDFs extracted.
Most retrieval failures get blamed on the embedding model. In practice the model is rarely the problem: the chunk it was handed never contained the answer, because the boundary cut the paragraph in half.
2,140 words extracted
It gets a real title, a short summary and tags, written by the AI you connected.
Why your RAG pipeline returns the wrong chunk
Most misses come from chunk boundaries, not the model.
It joins a related collection, or starts a new one if nothing fits.
AI engineering
It's indexed by meaning, so the words you remember later still find it.
why does my chatbot cite the wrong doc
Top result, matched by meaning
Why your RAG pipeline returns the wrong chunk
Search when you know what you're looking for. Ask when you don't.
Search by words, topics, or names—or describe what you remember in plain English and let Savedly find the relevant memories for you.
Answers are written from your own saves, and every one lists the memories it came from.
What did I save about vector search?
Searched your memories for “vector search” · 3 found
You saved two pieces that cover this.
HNSW index tuning in Postgres says raising m and ef_construction buys recall at the cost of build time.
Why your RAG pipeline returns the wrong chunk blames chunk boundaries, not the embedding model.
Sources
HNSW index tuning in Postgres, explainedpgnotes.ioWhy your RAG pipeline returns the wrong chunkfieldnotes.devThe exact word or just the gist. Results found either way are ranked together.
HNSW index tuning in Postgres, explainedWebsite · pgnotes.ioKeywordMeaningWhy your RAG pipeline returns the wrong chunkWebsite · fieldnotes.devMeaning
Designing calm softwareVideo · vimeo.comMeaningSee the connections between the things you've saved. Hover a save or a collection to explore.
Anything in the vault disappears from search, the graph and sharing until you enter your PIN. It's PIN-protected, not encrypted.
Live: switch tabs, then come back.
Q3 planning — reading list and goalsPDF · 12 pages
Lisbon — 4 days, 3 neighbourhoodsScreenshotConnect OpenAI, Anthropic, Google Gemini, Groq or OpenRouter and pick a model for each job. You pay your provider directly.
Articles and pages, with the text pulled out so you can search what they say.
Half-formed ideas and quick thoughts, filed alongside everything else.
Images are read, so the words inside them are findable too.
PDFs and files, extracted and summarized instead of buried in a folder.
Saved once, then searchable by what was actually said in them.
Your memories should belong to you. Savedly is open source, so you can see how it works, run it yourself, and keep control of your data.
Self-host it · Read the code · Contribute
The code is public. Read it, fix something that bugs you, or build something it's missing. Every contribution, small or large, is welcome.
Fork the repo
Fork it to your own GitHub account and clone it locally.
Make your change
Create a branch, fix the bug or build the feature, and commit it.
Open a pull request
Push your branch and open a PR describing what changed and why.
The honest version of what the rest of this page claims.
Links, notes, screenshots, PDFs and videos.
You don't have to remember where you saved it. Just ask Savedly when you need it back.
