The historian energy companies would build for themselves.

Status
In development
Pilots
Seeking pilot operators for 2027
Priced
By the well or the facility, not by the tag
Posture
Read-only. It never writes to control

Why now

Fewer people are covering more wells. More of the decisions are getting made away from the pad. Work that took an engineer a week is starting to be done by a model in an afternoon.

Every one of those runs on history — complete, in context, and reachable. That is the part that has not kept up, and the tools built for the way the business used to run are not going to close the gap.

Open protocols and commodity storage mean it can finally be built the simple way, and built to work with you rather than around you.

Easy to stand up

Point it at a broker, let it find what is already publishing, and be storing history the same day. Not a six-month integration.

Low maintenance on purpose

Open storage formats and no proprietary catalog to fall out of step. It should run quietly for years without a specialist on call.

Open at both ends

Standard protocols in, and back out the same way — MQTT, Kafka, open formats and an MCP server. It works with the tools you already run.

Built for you, not against you

Priced by the well or the facility, read-only into control, and a maintenance agreement you can read in one sitting.

Built with operators

Energy is critical infrastructure. The companies that run it should have a say in the software that holds their history.

Twenty-five years in the industry means we have sat on the other side of these decisions. So we ask. What an energy company tells us it needs is the input that decides what gets built.

We are not an IT shop doing IT for its own sake. We build technology so the people who produce energy can keep producing it.

That input is open right now. Tessarius is in development, which is the most useful moment to be told we have something wrong. If you run an operation, or you used to, write to me directly and it will go straight into what gets built.

What we build

  • A time-series store built for OT Real field scan rates and real retention, with compression tuned to how process tags behave.
  • An asset model treated as the product Robust enough for the way a real operation is actually shaped, and general enough that any company can model itself in it. The assistant proposes it from what is already arriving.
  • Real-time calculations and analytics Rates, efficiencies and rollups computed as data arrives and kept as history. Drafted by the assistant, accepted by a person.
  • Alarm intelligence A cascade of alarms correlated against the history and collapsed to the condition that caused them.
  • AI in every part of it Building the model, writing the calculations, watching the alarms, answering the questions.
  • The data is yours Standard protocols and open formats on the way out, not just in, with an MCP server included. Nothing is locked inside a proprietary format only we can read.

That list is not finished, and it is being written right now.

Tell me what we are missing

How it works

01 — It arrives

At the resolution you choose

Every value your edge publishes is there to keep, with its units and its range, so nothing has to be guessed at later.

02 — It is held

A dropped link loses nothing

A connection that goes down does not leave a hole you find out about months afterward.

03 — It is stored

Written once, never edited

What the instrument said stays what the instrument said. A correction sits beside the original.

04 — It is used

Rates as the data lands

Derived values are computed on arrival and kept as history, so they trend like anything else.

05 — It goes out

Back where you need it

Streams, an SDK, your own warehouse, or an agent — on the topics you choose.

Getting data out

Where the history needs to land is different for everyone, so there is more than one way out of Tessarius.

The tools you already run

Existing screens and visualization tools subscribe over MQTT or query the API. Nobody has to learn a new front end to look at their own data.

SDKs for partners building on it

A direct path to the history for anyone building a product on top of it, fast enough to sit behind a real application.

Into your lake or warehouse

Move history in bulk into whatever lakehouse, warehouse or data platform your analytics already runs on, in open formats.

Streams and agents

MQTT and Kafka topics for systems that want values as they land, and an MCP server for agents that want to ask.

Read-only by design

Tessarius listens. It does not control. There is no path in the product that reaches back into your control system and changes anything.

The asset model

A pressure reading is just a number until you know which well it came from. The model is built to hold any operation, however yours is actually shaped — not a template you have to bend yourself into.

One asset, counted several ways

A meter can sit on a route and inside a gathering system at the same time, without being built twice.

Reorganize without rewriting history

Move a well to a different route in March and February’s totals stay exactly as they were.

Equipment brings its specifications

Serial number, manufacturer, model and rated capacity, beside the measurements.

Your vocabulary, not ours

Well, pad, unit and route are configuration. Rename them, reshape them, or model something else entirely.

AI and agents

Agents do real work here. They propose your asset model, draft calculations and alarm rules, answer questions about your history in plain language, and watch for the patterns nobody wrote a rule for.

They can do that because the asset model tells them what every number is attached to. An agent that knows a reading came from a gas lift well on Pad 7 can answer a question about that well. One looking at a tag name cannot.

It proposes, a person approves

The assistant drafts hierarchies, mappings, calculations and alarm rules. None of it takes effect until somebody accepts it.

Ask in plain language

Questions put to your own history and your own model, answered with the assets named rather than the tags.

An MCP server, built in

Your own agents connect directly — to query history with the asset context attached, and to build on it: hierarchies, templates, calculations and alarm rules, through the same interface a person uses.

Agents are audited like people

What an agent did is in the same log as what a person did, and read-only applies to agents exactly as it applies to us.

How it is priced

An E&P company counts wells. A midstream company counts facilities. We bill against whatever you already count.

By the well, or by the facility

You are billed against the unit your business already runs on, and the number on the invoice is one you can forecast.

Instrument it however you like

Add sensors, raise scan rates, keep the history longer. None of that changes what you pay.

An operations decision

Whether to measure something should be answered by the people running the field, not by a license.

Your unit, defined once

The billing unit comes out of your own asset model, so what you are charged for matches how you already report.

Where we are

We start in upstream oil and gas, with operators standardizing on Ignition, taking data in over MQTT Sparkplug B. The same architecture carries into the rest of energy, and into any industry that runs on real-time operational data.

Tessarius is in development. The ingest path runs end to end on a test bench today, and the administration interface is designed and in review.

We are looking for a small number of pilot operators for 2027. If that could be you, we would rather talk early than late.

ben.manek@tessarius.com