file organizer

Drop it. Find it.
Forget filing.

embly reads every file you drop in, classifies it, embeds it, and files it where it belongs. Search in plain English. Never open ~/Downloads again.

Open appWatch walkthrough
embly - inboxready
embly
work
Inbox
Explorer
review
Review
Suggestions
Inbox
drop any file to begin
Recent
saas-playbook.pdf
work/resources/books/the-saas-playbook
document2.2 MBApr 10
cup.py
technical/code-snippets/python
document1.1 KBApr 10
cup.jpg
other/stock-images/kitchenware
image74.9 KBApr 10
ssl-related.txt
technical/security/ssl-certs/production
document105 BApr 10
ingestextractclassifyembedfile
01 · the problem

Filenames are a terrible index.

14,392
files in your Downloads folder.
screenshot-2024-03-17-FINAL (3).png
is somewhere. You won't find it.
Search by filename
is how computers worked in 1998.
02 · how it works

Four steps. One pipeline.

01

Ingest

embly.watch(~/Downloads)

Drop, drag, or point it at a folder. Watches in the background.

02

Classify

llm.tag(content, context)

A local model reads the file and assigns tags, category, and a destination path.

03

Embed

vec = embed(chunks)

Content is chunked and embedded into a local vector index. Stays on your machine.

04

Search

find("tax pdfs last month")

Ask in plain English. Filter by type, date, or folder. Instant.

04 · features

Built for files you actually have.

"api key rotation runbook"
MDsecrets-rotation-runbook.md0.94
01

Semantic search

Query by what files mean, not what they're named. Local embeddings, cosine similarity.

invoice-mar.pdf
124 KB · PDF
#finance#invoices#work
Work/Finance/2025/
02

Auto-classification

Every file gets a category, tags, and a destination path. Override any — embly learns your taste.

no cloudno accountworks offline
~/.embly/vault
4.2 GB / 12 GB
2,418 files
2,418 indexed
84 tags
03

Local-first vault

Index, embeddings, and originals live on disk. No cloud round-trip. No account required.

.pdf.mdimageafter:jan
contract-2025.pdf.pdf
photo_001.jpgimage
tax-return.pdf.pdf
147 files — saved as "tax '25"
04

Type filters

Narrow by extension, MIME type, content kind, or tag. Chain them. Save as smart views.

PDFattention-paper.pdf
MDembeddings.md
IMGphoto.jpg
⎵preview
attention-paper.pdfp.1 / 14
05

Inline preview

Space to peek at anything in the result list: PDFs, code, images, markdown, transcripts.

quarterly-report.pdf
hero-image.png
meeting-notes.mdqueued
watching ~/Downloads
06

Drag-and-drop ingest

Drop a file, a folder, or a hundred. Watches folders in the background and processes live.

05 · built for

People with too many files.

/devs

Your screenshots folder has 3,000 entries. One of them is the auth bug from last sprint. embly finds it.

— git diff — welcome back
/researchers

Papers, notes, half-read PDFs. Ask for the one about attention from 2017. Not the filename — the idea.

— bibtex not required
/freelancers

Invoices, contracts, receipts — sorted the moment they hit disk. Tax season stops being a search party.

— April, unbothered
06 · questions

The short answers.

On your disk. The vault, index, and model weights all sit under ~/.embly. Nothing leaves the machine unless you point it at a remote model.
PDF, plaintext, markdown, HTML, DOCX, RTF, EPUB, CSV, JSON, most code, images (OCR + vision), and audio/video transcripts via Whisper. Anything else is indexed by filename and metadata.
Yes. The default model bundle runs locally — classification, embedding, and search all work with no network. Remote model providers are opt-in.
On our internal benchmark of 40k mixed files: 91% top-1 category, 97% top-3. You can correct any decision and the system adapts to your folder structure within a few edits.
TBD. The CLI and local engine will stay free. A hosted sync layer is planned for teams who want shared vaults — pricing when we ship.