Your private data agent.
Not a warehouse.
Sery puts AI to work on the documents that can't leave your office — patient charts, client contracts, student records, tax workpapers. Read, searched, summarised, and reasoned about on your machine. Not the cloud.
Our mission
Every clinic, firm, and school is sitting on a pile of documents that AI could help with — patient charts, client contracts, student records, tax workpapers — but those files can't go to a cloud LLM. Privacy law, client confidentiality, plain professional caution. So today the AI productivity revolution skips the regulated half of the economy.
Sery closes the gap. The desktop agent reads PDFs (incl. scans via OS-native OCR), DOCX, PPTX, XLSX, CSV and Parquet on your machine via an in-process Rust toolkit. AI finds, summarises, drafts, compares, and reasons across them — and the bytes never leave the storage they live in.
Once you install Sery on more than one machine, it stops being an app and starts being a private workspace — the office iMac, the partner's ThinkPad, the home laptop, and every cloud bucket they each connect to, all readable by AI as one. Without any document ever crossing the wire.
An agent, not a warehouse
Sery's cloud is a coordination layer, not a storage layer. We don't hold your documents. We hold a small, content-blind catalog — filenames, schemas, column statistics — that lets the agent route a question to the machine where the answering document already lives.
Your private files never leave the machines they live on. When you ask a question, the answer crosses the wire — the document doesn't. When you turn the cloud off, every local feature still works. The structural property is the product.
That means our category isn't cloud storage and it isn't a hosted AI tool. It's a private document agent for the files that already live on your machine.
Why we built this
The specific moment: a friend who runs a CPA firm, describing tax season. Three hundred clients, every one of them a folder of PDFs — 1099s, K-1s, brokerage statements, prior-year returns. Every junior burning hours opening, reading, transcribing, summarising. Cloud AI sitting right there, useless. None of those documents can go to it.
The AI productivity revolution was skipping the half of the economy that actually deals with private documents.
Every existing tool missed the mark:
- ✗Cloud AI tools: powerful, but the documents that need them most can't be uploaded — patient charts, client contracts, student records, tax workpapers.
- ✗OS-level file search: knows filenames, not what's inside the PDF. One machine at a time, never across cloud storage.
- ✗Cloud BI tools: want you to upload everything first. Your files aren't a database, and you don't want them on someone else's server.
- ✗Cloud storage browsers: show you the file list but make you download everything just to see what's inside. No AI. No search across buckets.
We thought: “What if AI could read every private document without any of them leaving the machine they live on?”
That question became Sery Link. No warehouse. No upload. No setup. Install it, connect your storage, and AI gets to work on the files that can't leave the office.
How we're different
Reads PDFs and scans on your machine
A native Rust extraction toolkit (mdkit) reads PDFs via pdfium, scanned pages via macOS Vision and Windows.Media.Ocr, Excel via calamine, and DOCX/PPTX/EPUB/RTF/ODT via the Pandoc sidecar — all in-process on your laptop. Patient charts, client contracts, and tax workpapers become AI-readable without ever leaving the device.
Find / Summarise / Draft / Compare / Reason
Five verbs, one private workspace. Find indemnification clauses across a folder of contracts. Summarise this morning's patient intakes. Draft personalised feedback from a stack of student essays. Compare last quarter's brokerage statements. Reason about cohort eligibility across charts.
Every store and database your data lives in
16 sources: 9 file stores (Local, HTTPS, S3 + B2/Wasabi/R2/GCS presets, Google Drive, SFTP, WebDAV, Dropbox, Azure Blob, OneDrive) and 7 databases (Postgres, MySQL, Snowflake, ClickHouse, MongoDB, Redis, SQLite). Add a source in seconds — credentials in your OS keychain, not our servers.
Content + filename + column search
Nothing else understands that a PDF has paragraphs and a CSV has columns at the same time. Search 'indemnification' and you get hits across contracts AND clause-named columns. Search 'price' and you get every tabular file with a price-like column — even when the filename says nothing about it.
AI in the cloud, documents in your storage
Connect a Sery workspace ($19.99/mo) and ask the agent at app.sery.ai. The agent runs server-side; reading and execution travel to the machine where the document lives via a private tunnel; only the answer comes back. Raw files never cross the wire.
Small and unobtrusive
Lives in your menu bar. Auto-updates. macOS, Windows, Linux. Watches folders + remote sources in the background; wakes up when you need it.
Open on the desktop
The Sery Link desktop app is open source under AGPL-3.0. Every line of code that touches your files lives on github.com/seryai/sery-link. Read it, fork it, audit the network calls — the privacy guarantee is structural, not promised.
The cloud coordination layer is the business. Free forever for unlimited sources and unlimited machines, fully offline, no account required. $19.99/mo unlocks AI chat across all your sources, multi-machine search across your workspace, and lets you share with up to 5 invited machines on one bill — for a teacher and class, a consultant and clients, or a household.
Who it's for
Anyone whose most useful documents can't go to a cloud LLM. Clinicians with patient charts. Lawyers with client contracts and discovery. Accountants with tax workpapers, K-1s, and prior-year returns. Teachers with student essays and gradebooks. Plus the developers, designers, and power users with private files scattered across S3, Drive, Dropbox, SFTP, and a NAS at home.
If your current strategy is “skip AI on this one because it can't leave my machine” — Sery is for you.
Put AI to work on the documents that can't leave.
Free, open source, runs on your machine. Reads PDFs, scans, DOCX, XLSX, CSV, Parquet — all in-process. No account required to install.