Explore AgentFEM¶
An engineering question can become a model, a collection of simulations, or an interactive design tool. These three starting points show how those pieces fit together. Choose the one closest to what you want to do today.
See the design decision¶
How much material can a support lose while limiting its deflection?
Rotate a three-dimensional support, compare alternatives under the same load, and inspect the weight–stiffness trade-off in the AgentFEM × GINO structural-design lab. The browser demonstration requires no installation.
Give an AI agent a real simulation workflow¶
Can an agent operate a finite-element model without hiding the model from you?
The AI-assisted finite-element quick start follows one cantilever from project creation to an inspectable result. The same project is ordinary Python, whether a person or an MCP-compatible agent operates it.
Turn simulations into learning data¶
How do repeated finite-element runs become a surrogate model?
The simulation-to-surrogate walkthrough varies one material property, solves the corresponding models, and trains a compact baseline. It is deliberately small enough to understand before scaling to geometry variation, field prediction, or neural operators.
Keep exploring¶
- Executable example catalog: elasticity, heat transfer, dynamics, material behavior, and public benchmark comparisons.
- Installation: Linux, macOS, and Windows through WSL2.
- AgentFEM-Learning: optional scientific-learning providers and their current evidence.
- Source and community: code, issues, releases, and discussions.
- Preserved source snapshot: Software Heritage's completed archive of the repository, captured on 19 September 2026 (UTC). Use the source link above for current development.
AgentFEM is Apache-2.0 licensed and builds on FEniCSx/DOLFINx, PETSc, and MPI. The examples show concrete workflows, not a claim that every engineering problem has already been qualified.