
ChemE-as-Code: an open-source platform for agentic AI in chemical engineering
We release ChemE-as-Code, an open-source platform on which AI agents can carry out chemical engineering work and engineers can check it. The accompanying paper is available as a preprint.
In ChemE-as-Code, a plant is kept as linked, machine-readable engineering files in Git. Calculation tools, from catalysis through process simulation to plant layout, HAZOP and life cycle assessment, run outside the language model, and skills tell a general-purpose agent how to use them. Each value records where it came from, and automated checks report inconsistencies and out-of-date results by file and line, so that engineers can see what to review.
In the paper, agents develop a natural-gas ammonia plant from a written task to a preliminary basic-engineering package with simulation files, P&IDs, a 3D layout, a HAZOP, datasheets and relief cases, for USD 224 of API cost in the design sessions. Planted faults and engineering changes show that the checks find what their rules cover and miss the rest: they guide human review rather than replace it. No independent engineer has reviewed the design.
The platform comes with the Process Atlas, an open library of about 50 chemical processes kept as code, with life cycle inventories for Brightway, openLCA and other tools.
- Paper: Artur M. Schweidtmann, ChemE-as-Code: An Open-Source Platform for Agentic AI in Chemical Engineering, preprint, submitted to AIChE Journal. Download the preprint (PDF)
- Project website: cheme-as-code.org
- Software page: ChemE-as-Code