Vision
A new way to do chemical engineering
ChemE-as-Code treats a chemical plant like software: open, machine-readable files that engineers and AI agents check, reuse and improve together. We believe this will change how our whole field works.
From documents to code
Chemical plants are still designed in drawings, spreadsheets and documents that drift apart, locked in proprietary tools and company archives. Every project starts largely from scratch, checks are manual, and knowledge leaves with the people who had it.
As code, a design becomes a set of plain files in version control. Tools check that the P&IDs, process model, equipment and life-cycle data agree; every change is a readable diff with a reviewer; proven designs become building blocks for the next project; and AI agents can read, check and extend the same files as engineers do. That speeds up development and changes the way we work, from the first sketch to the life-cycle assessment.
Open by default
Process design data has largely been proprietary: kept in company archives, sold in licensed databases, or reduced to a few results in a paper. The Process Atlas is an effort to make it openly accessible: the code under the GNU AGPL-3.0-or-later, the process data under CC BY 4.0. Open data means fairer access to information, for students, researchers, small companies and regions that cannot pay for expensive licences, and research that others can verify and build on.
Every detail, not just the number
Life-cycle inventory databases such as ecoinvent publish inventories, mostly under licence and without the engineering behind them. The Atlas publishes the design an inventory comes from: flowsheet, P&IDs, equipment, operating conditions, the assumptions and their sources, and how far the design has been taken and checked. So you can trace every number, adapt the design to your case and compute the inventory again.
Every change on record, with credit
The library is a git repository. Every change to a process is kept with who made it, when and why, and can be compared with any earlier version. Reviews happen on that change, in the open. Nothing is overwritten silently, and every contributor is credited for their work.
Machine-readable, for engineers and AI
Everything is in open, standard formats: DEXPI for P&IDs, plain YAML for process models and data, open inventories ready for Brightway and openLCA. The files are interoperable, findable and reusable. An engineer can start a new design from an existing process; an AI agent can read it, check it and propose changes, with the same tools validating the result.
A community effort
No single group can describe the chemical industry. The Atlas grows and stays correct through its users: report an error or ask for a process in an issue, suggest an improvement, or add a process of your own.
Contributions go through the repositories on GitHub.
Ideas behind it
ChemE-as-Code is an open implementation of Plant-as-Code, proposed in a Nature Chemical Engineering Comment (Schweidtmann & Demeyer, 2026), which it extends beyond the plant to chemistry, experiments and life cycle assessment. It applies Engineering-as-Code, engineering design kept as versioned text with automated checks after the example of Infrastructure-as-Code, to chemical engineering. Changes follow a Design-Ops workflow: a project moves through stages, each with gates that are checked and sign-offs that people give.
Everything published is research output, not engineering advice: see what this means.