The managed data box for machine fleets

Your databases, answering questions.

Directly in Claude and in Excel — from one dedicated box we run for you. No data warehouse, no BI rollout, and your data never pools in anyone’s SaaS.

  In production today

Running a real fleet, right now.

3,900+
machines under analytics, across 40+ tenants and 2,900+ stores
4
continents of fleet data — one box
550K+
telemetry rows refreshed hourly, published with zero-downtime hot swaps
1,800+
questions answered by business users through Claude and Excel in the first production month

The numbers above are one customer: a global beverage-equipment operator whose leadership KPIs, fleet health, and per-machine diagnostics all run through a single IOT Data Flow box. Built for operators of distributed machine fleets — beverage & vending, EV charging, laundry, HVAC. If your machines write to MySQL or Postgres and your leadership asks its questions in chat, this box is for you.

How it works

One box. Three doors.

Instead of a dashboard rollout, an export pipeline, and a BI seat for every person — each audience gets a scoped door into the same governed box.

01 — Leadership & partners

Ask in Claude

“How many drinks did we sell yesterday?” — answered in plain language from live data, through curated KPI datasets and guard-railed tools. Never raw tables, unless you say so.

https://you.iotdataflow.com/mcp/<token>
02 — Finance

Refresh in Excel

Live figures straight into Excel and Power Query. Refresh the workbook, get today’s fleet — no exports, no copy-paste.

…/query?format=csv&token=<token>
03 — Engineers & analysts

Query directly

A read-only SQL console, full REST API, and MCP for coding agents — across every connected engine.

GET /api/sources · POST /query
Less, but better

A single hardened binary.
Nothing else to operate.

One box per customer

Physically isolated on a dedicated server — never multi-tenant. Your data stays yours.

Curated datasets & guides

We shape raw tables into KPI views with plain-language guides, so the AI answers in your business’s vocabulary — not in column names.

Agent tools

Pre-built, parameterized questions — “is machine X’s temperature normal?” — with values safely bound. The AI presses buttons you designed; it doesn’t freestyle SQL against raw data.

Read-only by design

A query guard blocks writes; per-query timeouts and row caps keep any one question from hurting a database.

Scoped, revocable access

Business links reach only what you’ve shared. Every query is logged — who, when, what, and how many rows.

Config-as-code

The box’s whole configuration exports as one reviewable YAML manifest — versioned nightly, restorable anywhere.

Why not just point Claude at your database?

More than a connector.

A raw MCP connector wires one developer’s Claude straight to one database, with whatever credentials you hand it. IOT Data Flow is the governed layer you put in front.

A raw MCP connector
IOT Data Flow
Who it serves
One developer’s Claude session
Leadership in Claude, finance in Excel, engineers in SQL — one box
Access scope
Whatever credentials you give it — often the whole database
Scoped doors: curated-only or full, chosen per audience
Credentials
Live in each person’s local config
Stay encrypted on the box; people get revocable links
Safety rails
Whatever that one server happens to enforce
Read-only guard, timeouts and row caps, enforced centrally
Governance
None central — no audit, no revoke, no curation
Full audit log with row volumes, curation, nightly encrypted offsite backups

IOT Data Flow speaks the same MCP — it is a connector. It’s just the one you can safely give your whole company.

Security

The strongest isolation is physical.

  • One dedicated box per customer — never multi-tenant.
  • The box pulls with read-only credentials. Nothing is installed in your network, and nothing connects inbound to it.
  • Business links reach only curated datasets and approved tools — raw tables stay locked away.
  • Every query logged: who, when, the exact SQL, and how many rows.

Read the security architecture →

Written for your security team: the full architecture on one page — connectivity options, scoping, audit, what the AI can and cannot do, EU hosting, and two-layer disaster recovery. Plain language, no diagrams-as-decoration.

From credentials to answers

Live in a week.

Day 1

Provision & connect

We stand up your dedicated box — EU hosting or your own cloud — and connect your databases with read-only credentials. No agents installed, no inbound access to your network.

Week 1

We curate with you

Together we turn raw tables into guides, agent tools, and KPI datasets: the questions your team actually asks, encoded.

End of week 1

Your team asks

Leadership pastes one link into Claude; finance refreshes Excel. You watch every question arrive in the audit log.

  Get started

See your fleet answer questions.

Bring a read-only credential; leave with your leadership asking your live data questions in Claude. We’ll show you on your data, not a demo dataset.