Why we built Sery
OS-level file search knows filenames. Your files contain columns. Sery Link is the file manager that indexes both — plus the content of your documents — so you stop opening five CSVs to find the right one.
The Monday morning file hunt
A partner at a small accounting firm spends about forty minutes every Monday looking for a file. She knows it exists. She knows she worked on it last quarter. She knows it had a column called vendor_id — because that's the whole reason she needs it again. What she doesn't know is which of the forty-something exports sitting in her Downloads folder it actually was.
Her options, all of which she has tried: open each file in a spreadsheet app (which hates a hundred-open-tabs). Use the built-in OS file search, which searches filenames and a bit of content but has no idea that the file has a vendor_id column. Keep a sticky note of which file has which columns, which lasts until the next week of work invalidates it. Re-export from the source system, which is slow and somehow always produces slightly different column names.
This is not a one-person problem. It's what it feels like to work with data-heavy files in 2026. Every existing tool misses the mark because the tools were built for a different unit of organization than the one most people actually think in.
What we think is actually going on
The software that indexes your filesystem — every OS-level file search built in the last twenty years — operates at the file level. It knows filenames, maybe extensions, sometimes the text body of a document. It does not know that a CSV has columns, which is almost always the thing you actually remember about it.
The software that understands columns — cloud BI tools, warehouses — operates on data that's been uploaded to a server first. It won't look at the CSV in your Downloads folder, and the friction of getting it there is what makes most people give up and just open files one by one.
Between the two there's a gap. A file manager that understands what's inside the files it's showing you. That's what we built.
Column-aware search, on your laptop
Install Sery Link on a machine, point it at a folder, and it walks the tree. For every CSV, Parquet, Excel file, and document it finds, it extracts the schema (column names + types) using an embedded SQL engine locally, pulls a handful of sample rows, and writes the result into a small index that lives in your home directory. This runs in the background; a file watcher picks up changes.
Then there's one search bar. Type a filename, it's a filename search. Type a column name like vendor_id or price and Sery hits every tabular file with that column — even if the filename says nothing about it. Type a word that's inside a DOCX or PPTX and you get content matches too. Each result tells you why it hit, so you can see at a glance whether you matched on filename, a column, or some text inside.
None of this requires uploading anything. The index is a local SQL file on your laptop. Your data stays on your laptop. The app works fully offline from install through daily use.
The same local engine also reads remote sources. Connect an S3 bucket, Google Drive, SFTP server, WebDAV share, Dropbox, Azure Blob, Cloudflare R2, Backblaze B2, Google Cloud Storage, or OneDrive — Sery fetches only metadata and sample rows for each file, never a full download, and indexes them alongside your local folders. Nine source protocols and four S3-compatible presets in one search bar.
AI is an upgrade, not a dependency
Everything above is true with zero cloud contact. When you want more — natural-language queries like "total revenue last quarter across my spreadsheets" — you have two paths.
One: connect to sery.ai with a workspace key. Our cloud agent fans SQL out to your machine through a private tunnel. The query executes locally; only the result rows travel back. Raw files never leave your storage.
Two: run Sery Link with --mcp-stdio and point Claude Desktop, Cursor, or any MCP-capable client at it directly. No cloud account. The model talks to your local index over the Model Context Protocol. Your editor gets column-aware search and SQL execution as native tools.
Either way, raw file bytes stay where they live. We think of this as the inversion of how most AI data tools are built — they start from the cloud and graft local access on awkwardly. We started from the laptop and let AI be the natural next step when you want it.
Who we built this for
The first user of Sery Link is someone whose day-to-day involves a lot of data files they keep losing track of. Finance people with exports from QuickBooks, NetSuite, Stripe. Analysts juggling client CSVs. Scientists with datasets sprawled across a laptop and a NAS. Journalists digging through FOIA dumps. Accountants during tax season. Founders before anyone on the team has built a real data platform.
If your current strategy for finding a file is "I know it's in one of those folders" — Sery Link is for you.
Where we are now
Sery Link v0.8 ships column-aware search, per-file column profiles (null %, unique values, min/max/avg), a file detail view, nine remote source protocols (local, HTTPS, S3, Google Drive, SFTP, WebDAV, Dropbox, Azure Blob, OneDrive) plus four S3-compatible presets (B2, Wasabi, R2, GCS), an MCP server for direct AI client integration, and a fully offline mode. The free tier has no account requirement — connect to sery.ai only when you want AI queries or multi-machine search.
If you try it and have thoughts, we read every message at [email protected].
— The Sery team