Simulators Overview
On top of its general agent capabilities, ChemAgent ships a chemical engineering layer: the agent drives a locally installed process simulator to build models, run cases, and read results, and coordinates dedicated optimization design modules. This layer is what separates ChemAgent from a general-purpose assistant.
What it does
Describe the task in natural language; the agent defines components, selects property methods, builds columns and exchangers, connects streams, and runs the case to convergence.
After a run the agent reads tray profiles, stream compositions, and energy figures, and explains them in the conversation.
Column-specific calculation modules return a design option in seconds, and the approved option carries straight into simulator modelling.
Simulation conclusions can be stored in the knowledge base, and process constraints from the knowledge base inform model inputs.
Registered on demand
- ✓Appears only where installed
Simulator toolkits register only on machines with the simulator and the licensed computation components; every other capability works without them.
- ✓Validated before execution
Tool inputs are checked before anything reaches the simulator, so malformed parameters fail fast with a clear message instead of corrupting a model.
- ✓Per-toolkit switches
Each toolkit appears as its own group in the console skills page, with individual tool switches.
- ✓Mode-gated
Simulator tools are exposed only in Build and Auto modes; Default and Knowledge conversations cannot touch a simulator.
Exclusive session guard
A process simulator is a single-instance program, so ChemAgent enforces first-come ownership: the first session to use a simulator holds it, other sessions are told who holds it, and closing the simulator releases it. Concurrent sessions cannot corrupt each other's models.