This is not EPANET.

1. We are deeply dependent on EPANET

Not just inspired by it. Not just compatible with it. Dependent on it. The specifics, each one listed with its source on the claims page:

• The engine is EPANET's, and it ships inside the page. A 679 KB browser build of OWA-EPANET 2.3.5, released 2025-02-20, is vendored in the repository and loaded on demand. We did not write a solver engine to compete with it. We wrote a drawing surface and put EPANET underneath it.
• Most of the answers only exist because it is there. Extended period simulation, water age, source trace, chemical decay, pump energy, rule-based controls, and pressure-reducing, pressure-sustaining, and flow-control valves all run through the EPANET engine and through nothing else. Our own solver does steady state (single moment), pressures and flows only. It has no time dimension and is not currently planned to get one.
• The interchange format is EPANET's. We read and write EPANET's own .inp file, section by section across the sections we support, from [TITLE] to [END], and a value you did not edit comes back out character for character. We also have a native format of our own: a project saves as JSON, because there are things an .inp cannot hold. Scenarios. Saved profile paths. The background picture itself, which an .inp can name but cannot contain. Map tiles and the view you were looking at. A Text label anchored to a pipe, or one with more than one line in it. Longitude and latitude for a project drawn on the world. Everything an .inp can carry still round-trips through one, and the five things it genuinely cannot are reported to you rather than silently dropped.
• We check ourselves against EPA's own published answers. Our extended period simulation is compared against EPA's own report for its Net3 example network: 2,425 head comparisons across 25 reporting steps, agreeing to 0.005 ft. Water quality is anchored on the same report, worst difference 0.105%. Pump energy is anchored on that report's own energy table. EPANET is not our benchmark by preference. It is the truth we are measured against.
• Even the arithmetic is EPANET's. Our one Hazen-Williams coefficient pair is derived from EPANET's h = 4.727 L Q^1.852 / (C^1.852 d^4.871), not from the more commonly quoted metric restatement, because EPANET is the program our users will check us against. The two disagree by up to 0.12% across pipe sizes from 50 mm to 2 m, and we chose to agree with EPANET.
• The vocabulary is EPANET's. Junction, reservoir, tank, pipe, pump, valve. Extended period simulation. Trials and iterations, static pressure, drawdown, converge. The four curve types. The grammar of a rule. What happens when you change a unit. An engineer who knows EPANET should not have to learn a second dialect from us, and we have a check that fails the build when we start writing one.
• The interface is inspired by and informed by EPANET's. Its curve editor and its table of assets are where ours came from, and its own five map colours are in our palette unsoftened, so a reader who knows an EPANET pressure map can have that one back, though our own shipped default is Viridis rather than EPANET's. There are some features that are not simply downstream: scenarios, a world map, find and replace, flexible multi-labels, and what we hope is a less confusing and easier interface (menu system). And it is a web page rather than a Windows program. Some of that is better and some of it is just different.

None of that is a complaint and none of it is modesty. It is the actual dependency graph. If EPANET had not existed, or had not been given away, this project would not exist.

2. Our licence is less free than EPANET's, and we should say so

Not public domain; less free than EPANET; less trusting than EPANET; arguably less collegial than EPANET. For better or for worse, we are currently working under Richard M. Stallman’s GNU General Public License 3.0, which is aggressively restrictive against locking up this work. The rationale is that this protects us all against property hoarders. Talk to us about this if you see the error of our ways.

The invitation at the end of that is real. EPANET carries no licence at all, because a work of the United States government has no copyright to license. Anybody may take it and do anything, including take it private. We chose a licence that forbids that, which is a restriction we impose and EPA did not. Reasonable people disagree about whether that is generosity or distrust. If you think we got it wrong, write to us.

3. We are new, with all that implies

We are not going to dress it up.

One thing is worth adding, because it is the claim about ourselves that a future surprise cannot falsify:

We are what we are, and it’s good and useful and usable. But we have no idea what we are not and what we don’t know.

We do not publish a completeness claim against EPANET, and we never will. EPANET has depth we keep discovering by accident.

4. This is AI-assisted, and here is the reason

The software and this website are written with heavy use of AI. That is a fact a user is entitled to know, so it is here rather than in a colophon. Tom Haws asked that his own statement of why be carried as he wrote it:

This would not be possible without AI. Tom Haws decided in 2026 that if AI is going to destroy the world, he better not use it for cat videos. His only justification for using AI is to tell the world the highest message of heaven: That they are loved and cherished forever, that they have nothing to fear, and that they are not ruining everything.

5. Nobody has vetted this on your behalf

No EPA affiliation, no government affiliation, no university affiliation, no sponsor, no foundation, and no funding. Nobody has vetted this software on your behalf. Every claim on this page is one you can check yourself: they are set out one by one, each with its source, on the claims page. Where we were not sure of a fact, we left it out.

The United States government, and the people who fund it

EPANET was paid for by American taxpayers and given to the world for nothing. Under 17 U.S.C. § 105, copyright is not available for a work of the United States government, meaning a work prepared by an officer or employee as part of that person's official duties. So EPANET has no owner to ask, no fee, and no permission to seek. It is in the public domain by operation of law.

That is an extraordinary thing to have done, and it is worth naming it as a gift rather than as a background condition. A working engineer anywhere on earth, in a country that pays no American taxes, can model a water system tonight because of a decision made in a United States federal agency.

The Environmental Protection Agency, and Lewis A. Rossman

EPANET was created by Lewis A. Rossman for the EPA's water research programme and first appeared in 1993, alongside a formal report on the EPANET water quality model. EPANET 2 followed, and version 2.2 was the last release EPA itself made. It is used by utilities, consultants, regulators, and universities worldwide, and its computational engine sits underneath a great many other products, ours among them.

We have never contacted the EPA and we are not asking them to notice us. The gratitude is stated here, publicly, because that is the only form of thanks a public-domain gift admits of.

Open Water Analytics

EPANET's development continued in the open after EPA's 2.2, as a community effort in collaboration with EPA. Open Water Analytics released EPANET 2.3 on 2024-07-17 and 2.3.5 on 2025-02-20. 2.3.5 is the engine our software actually runs. Every volunteer who has filed an issue or a fix in that repository has, without knowing it, fixed something for us.

The software and services we depend on

These are the third-party pieces the network editor uses. The first four are vendored, meaning a copy travels with the software rather than being fetched from somebody else's server while you work.

• The browser build of the EPANET engine. MIT licensed, © Luke Butler, wrapping the MIT-licensed OWA-EPANET engine and compiling it to WebAssembly. It is the single piece of code that makes an EPANET run possible inside a web page with no server, and it is the reason this project could be built by one semi-retired engineer rather than by a team.
• Bootstrap 5.3.2, MIT licensed, © 2011-2023 The Bootstrap Authors. The layout the calculator pages are built on.
• Colour schemes for maps, Apache License 2.0, © 2002 Cynthia Brewer, Mark Harrower, and The Pennsylvania State University. See the Cynthia Brewer entry below.
• viridis, magma, inferno, and plasma, released CC0 by Nathaniel J. Smith, Stefan van der Walt, and (for viridis) Eric Firing. github.com/BIDS/colormap

And two services the editor can call out to over the network. Both are off until you turn them on, each asks its own separate permission, and a tile request tells that service where you are looking:

• OpenStreetMap contributors, whose map of the world is the ground under a geographic project, and whose Nominatim service answers a place-name search. Map data © OpenStreetMap contributors.
• Mapbox, for satellite imagery and for the terrain elevation data a node can read its ground level from.

Cynthia Brewer, and the people who made maps readable

A network map is worth nothing if a reader cannot tell the classes apart, and a startling amount of the work of making that possible was done by cartographers rather than by engineers. Cynthia Brewer, a geographer at Penn State, published her colour scheme advice for cartography in 2002 with Mark Harrower and Penn State, funded by the National Science Foundation's Digital Government programme through the GeoVISTA Center. The schemes are designed per class count, are tested for photocopying, for projection, and for colour-blind readers, and they are the reason a five-class pressure map is legible at all. The International Cartographic Association awarded her the Carl Mannerfelt Gold Medal in 2023.

Her schemes ship in our software under the Apache License 2.0, whose clause 2 fixes the exact wording of the acknowledgement below and whose clauses 4 and 5 forbid using the project name to endorse or promote anything. So the sentence is reproduced verbatim and untranslated, and no control, menu, or feature in our software carries the ColorBrewer name. What follows is the licence's own text, not ours:

EPANET's own interface

It is customary in this field to describe EPANET's user interface as dated. We decline to. It is the interface that taught most working modellers what a network model is, and our editor is informed by it deliberately: the curve editor is modelled on EPANET's, and the table of assets is the view EPANET has had since version 2.0. EPANET 2.2's own five map colours are in our palette unsoftened, so that a reader who knows an EPANET pressure map can have exactly that one back. They are not our default, and it would be easy to imply otherwise: the shipped default is Viridis, chosen because it is perceptually uniform and readable to colour-blind viewers. The rest of the palette is Cynthia Brewer's.

Richard M. Stallman, and the Free Software Foundation

The Free Software Foundation released version 3 of the GNU General Public License on 2007-06-29. That licence is what our own software is published under, and it is the instrument by which what we publish stays published. We did not have to design a legal mechanism to keep this free. Somebody else spent decades doing that, and gave it away too.

Whether it is the right choice for a project this dependent on public-domain work is an open question, and honesty item 2 above is where we admit it.

And the people whose names we do not have

The engineers who published Net1, Net2, and Net3 as example networks, which is what anybody testing an EPANET-compatible program checks against first. Everyone who has ever answered a question about the file format in public. Anyone who has reported a bug to us, or corrected a translation, in any of 27 languages. If your work is in here and your name is not, that is our failing and we would like to fix it.