Sery Documentation

Goal: install Sery Link, pick a folder, search it. No account or signup needed for local features.

1. Install — download Sery Link for your OS and drag to Applications (macOS) or run the installer (Windows/Linux).
2. Open — the welcome wizard shows one screen: "Pick a folder."
3. Pick a folder containing Parquet, CSV, Excel, or documents you want to search. (You can also skip and add folders later from Settings.)
4. Sery Link indexes locally — no email, no signup, no cloud contact. Schemas, sample rows, and column stats are cached on your machine. Larger folders take longer to finish their first index.
5. Search anything — type a filename, a column name, or text from a document. Results assemble across all your watched folders.

Want to ask natural-language questions? Two paths:

• • Cloud workspace — connect to app.sery.ai for AI queries that route SQL across all your machines, with curated recipes for common workflows
• • MCP — let Claude Desktop / Warp / Cursor / Continue read your folders via local stdio (no Sery account needed)

Grab the build for your platform from the download page.

macOS

1. 1. Download the DMG for your chip (Apple Silicon or Intel) from /download.
2. 2. Double-click the DMG file.
3. 3. Drag Sery.app to the Applications folder.
4. 4. Open Sery from Launchpad or Applications.

xattr -dr com.apple.quarantine /Applications/Sery\ Link.app

$ brew install seryai/tap/sery-link

Windows

1. 1. Download the .msi installer from /download.
2. 2. Run it; accept SmartScreen if prompted (unsigned builds are flagged until we reach a reputation threshold).
3. 3. Sery installs to Program Files and adds a Start Menu entry.

Linux

1. 1. Download the .AppImage or .deb for your distro from /download.
2. 2. For AppImage: chmod +x sery-*.AppImage then run it.
3. 3. For Debian/Ubuntu: sudo dpkg -i sery-*.deb.

Local search works the moment you've added a folder — no AI provider needed. Type filenames, column names, or document text into the search bar; Sery Link returns hits with a "why matched" line so you know where each one came from.

Plain-English questions happen in your cloud workspace at app.sery.ai. Connect Sery Link to a workspace (paste a workspace key from app.sery.ai/settings/workspace-keys), then open the dashboard's chat — Sery's AI agent generates SQL, runs it on the right machine via the secure tunnel, and answers in plain English. Local-only users still get the file-manager surface; AI is the upgrade path.

Example questions

Once you've connected a workspace, things like:

• "Find my 2024 tax documents" — returns a file list with paths
• "How much did I spend on flights last year?" — aggregates matching CSV rows
• "What's my busiest folder by file size?" — ranks files across indexed data
• "Show me the largest PDFs modified this month" — filters + sorts

How the answer comes back

Sery responds with a short natural-language summary plus a structured answer — a file list for "find" questions, a table + optional chart for "how much / how many" questions.

Install Sery on every machine that has data you care about — home PC, office laptop, server, NAS — then connect them all to the same workspace to query across them. Machines are connected using a workspace key you create once on the Sery dashboard; the same key works on any number of machines.

Create a workspace key

1. 1. Sign in to app.sery.ai → Settings → Workspace Keys.
2. 2. Click "Create Key". Give it a name (e.g. "My Machines"). The key looks like sery_k_… — it's shown once, so copy it.
3. 3. The same key can connect as many machines as you want. Rotate or revoke it anytime from the same page.

Connect a machine

1. 1. Install Sery on the new machine (see Installation).
2. 2. Pick a folder (or skip) to complete first-run. Sery Link runs fully local at this point — no cloud contact yet.
3. 3. In the top status bar, click Connect next to "Local only".
4. 4. Paste the workspace key into the dialog. Sery Link stores the resulting token in the OS keyring; the key itself isn't persisted on the machine.
5. 5. The machine appears in the Machines tab on every other machine connected to the same workspace, and in the dashboard's Machines view.

Two pieces ship today, one piece arrives in v0.6. Both need a workspace connection — see Adding Another Machine above.

Cross-machine search (shipped today)

At app.sery.ai you can search filenames + column names across every connected machine in your workspace. Results include a machine column so you know where each file lives. Catalog metadata (paths, columns, row counts) is the only thing that crosses the network — files themselves never leave the device that owns them.

Cross-machine SQL via MCP (shipped today)

On the Plus plan, point an MCP-aware LLM client (Claude Desktop, Warp, Cursor, Continue) at https://mcp.sery.ai/sse and the LLM gets a query_sql tool that routes execution to whichever machine owns the file. SQL runs locally; only result rows come back. See the MCP Integration section below.

Cross-machine AI fan-out (v0.6 — not shipped yet)

Asking a single natural-language question that automatically dispatches to every machine in parallel and merges results — that's on the v0.6 roadmap. Today, the in-app Ask tab queries one machine at a time. Use cross-machine search (above) to find files across machines, then ask follow-ups against a specific dataset.

MCP (Model Context Protocol) lets external LLM clients — Claude Desktop, Warp, Cursor, Continue — read your data through Sery. Two flavors:

Local stdio mode (Free, no account)

Toggle a folder under Settings → MCP in Sery Link. Copy the generated config snippet for your client. Paste, restart, done. Nine read-only tools — list files, search, schema, sample rows, document text, SQL — all running locally.

{
  "mcpServers": {
    "sery": {
      "command": "/Applications/Sery Link.app/Contents/MacOS/sery-link",
      "args": ["--mcp-stdio", "--root", "/Users/me/Documents"]
    }
  }
}

Or skip the GUI: cargo install sery-mcp gives you the same MCP server as a standalone Rust binary.

Cloud endpoint (Plus only)

For multi-machine workspaces, point your MCP client at https://mcp.sery.ai/sse with a workspace API key (generate one at Settings → API Keys). Same nine tools, but query_sql routes to the machine that owns the file via the WebSocket tunnel.

{
  "mcpServers": {
    "sery-cloud": {
      "url": "https://mcp.sery.ai/sse",
      "transport": "sse",
      "headers": { "Authorization": "Bearer sery_xxx" }
    }
  }
}

Setup walkthrough + troubleshooting: MCP runbook.

Supported file types

Adding folders

1. 1. Open the main window → Folders tab.
2. 2. Click "+ Add Folder".
3. 3. Pick the folder. Sery watches it for changes so new files are indexed automatically.

You can add as many folders as you like. Sery indexes metadata only — raw file contents never upload, even when the folder contains large datasets.

Cloud sources

Sery Link can also pull data from cloud storage and shared URLs. Click "+ Add remote source" in the Folders tab to open the picker. Credentials and OAuth tokens live in your OS keychain on the desktop machine — they never reach Sery's servers, the same way local folders work.

Drive types we don't support yet: Google Docs, Forms, Drawings, and Sites. Sery Link tells you up-front how many were skipped when you watch a folder. Docs export support is on the v0.7 roadmap.

Here's exactly what leaves your machines, so you can audit it yourself:

What does travel

• • File paths, schemas, column names — so Sery can route your question correctly.
• • A small sample (up to 5 rows per dataset) to ground the AI's answer in real values. PII-looking columns (email, phone, SSN, tokens, etc.) are auto-redacted before they leave.
• • Your question, routed to each relevant machine.
• • Result rows — the final answer rows (capped at 10,000 per query).

What never travels

• • Raw file contents. Your files stay on the machines where they live. Queries execute locally; only the results come back.
• • Files outside watched folders. Sery only reads directories you explicitly add.
• • Redacted column values. Anything matching the PII heuristic is replaced with <redacted> before sync.

Sery has two plans. That's it. Full details at /pricing.

No Team or Enterprise tier yet. Sery is a personal tool in v1 — team pricing (SSO, audit logs) arrives once enough small firms ask for it. Cancel anytime from account settings.