This server has a remotely hosted endpoint. Open its connector listing for connection details, authentication requirements, and setup instructions:
Real finite element analysis from a conversation. An AI assistant sends a STEP file and a plain-English question ("does this arm survive a 6 g maneuver with a 400 g motor?"); this MCP server maps the words onto geometry, rejects bad setups deterministically before any compute is spent, meshes and solves with CalculiX on Modal as an async MCP task, and returns a verdict backed by numbers — peak von Mises stress, factor of safety, deflection, natural frequencies — plus rendered stress plots and an interactive 3D viewer. Every study ends with a non-skippable reaction-balance check; every benchmark on the validation page is regenerated by CI on every commit.
Measured on the live deployment (2026-07-31): cold start 3.8 s · cantilever study submit→completed 9.3 s warm / 12.9 s with a cold solver container · ~$0.002 per study at Modal's posted rates.
Benchmark | Reference | Computed | Error |
Cantilever tip deflection (PL³/3EI) | 0.58055 mm | 0.58046 mm | 0.02% |
NAFEMS LE10 σ_yy @ D | −5.38 MPa | −5.456 MPa | 1.41% |
Plate-with-hole peak (Heywood Kt) | 62.8 MPa | 61.92 MPa | 1.40% |
Cantilever f₁ (Euler–Bernoulli) | 816.0 Hz | 815.4 Hz | 0.08% |
Full table, mesh stats, and an explicit limitations section: docs/validation.md — regenerated by CI, never hand-edited.
Protocol: MCP 2026-07-28 (stateless core) with the
io.modelcontextprotocol/tasks extension implemented per SEP-2663 (the
official Python SDK v2 shipped without it — this repo also contains the
client-side extension that teaches the official
SDK to consume task-augmented servers) and an MCP Apps (SEP-1865) 3D
viewer. See docs/decisions.md for every engineering
call and why.
uv sync
uv run modal secret create mcp-fea-token MCP_FEA_TOKEN=$(openssl rand -hex 24)
uv run modal deploy -m src.server.modal_appModal prints the endpoint URL (https://<workspace>--mcp-fea-mcp-endpoint.modal.run).
Connect any 2026-07-28 MCP client with Authorization: Bearer <your token>.
First-time Modal users: uv run modal token new once. Then verify the live
deployment end to end:
MCP_FEA_LIVE_URL=https://... MCP_FEA_TOKEN=... make verify-liveRelated MCP server: abaqus-mcp
Tool | Sync/async | What it does |
| sync | STEP → face table (face_id, type, centroid, normal/axis, area, bbox) + part bbox/volume. Call first: it's how English maps to face ids. |
| sync | Built-in library: Al 6061-T6, Al 7075-T6, Steel 4140, Stainless 304, Ti-6Al-4V, ABS, PLA, Nylon 12, PC (E, ν, yield, density). |
| task | Linear static: validate → mesh (C3D10, curvature-adaptive, local thin-wall refine) → CalculiX → verdict/stress/FoS + 3 stress PNGs. |
| task | Natural frequencies + mode shapes ( |
| sync | Full payload for a completed task; |
Task lifecycle: working (statusMessage walks validating → meshing → solving → postprocessing) → completed | failed | cancelled;
tasks/cancel kills the running container. Tasks are retained 24 h.
One conversion boundary, at ingest. Internally everything is consistent
mm–N–MPa(–tonne): lengths mm, forces N, stress/moduli MPa, density tonne/mm³
(kg/m³ × 1e-12 — done once, in Material.density_tonne_mm3), which makes
modal frequencies come out directly in Hz. STEP coordinates are interpreted
in the declared units (mm|m|in); gmsh's own unit conversion is disabled
so the declared unit is authoritative — and sanity-checked (a 100 mm part
declared "m" gets flagged).
Every rejection returns {code, message, suggested_fix}.
Code | Meaning | Stage |
| STEP missing/unreadable/over 25 MB, bad units field | parse |
| CAD kernel crashed / hung on the file (isolated subprocess; endpoint unaffected) | inspect |
| Not in library / E ∉ [0.01, 1500] GPa, ν ∉ (0, 0.5), yield ≤ 0, density ∉ [10, 25000] kg/m³, missing density for modal | parse |
| Bad type; node pinned in two local frames | parse/deck |
| Bad type/units/direction; zero loads; bad n_modes | parse |
| Not a single solid body (assemblies = v2) | pre-mesh |
| Face id not on this part | pre-mesh |
| Constraints leave free motion (named, e.g. "rotation about the hole axis") — SVD rank check | pre-mesh |
| Load applied to a fully fixed face | pre-mesh |
| Part > 5 m or < 0.5 mm — declared units probably wrong | pre-mesh |
| σ ≈ F/A > 10× yield or < 1e-6× yield | pre-mesh |
| > 400k elements even after coarsening / local refine | mesh |
| Quality gates; thin walls get one local auto-refine first | mesh |
| Meshing exceeded its wall budget (near-tangent slivers, helical threads) — the worker is killed from outside | mesh |
| Mapped ccx failures (>150k elements auto-switch to the iterative solver) | solve |
| ΣRF vs free-node applied loads off by > 0.5% | post |
| > 5% of the load lands on constrained nodes (loaded face touches a fixed face) and is carried directly by the support | post |
| FoS < 1 (post-yield numbers flagged non-physical) / u > 5% of min dimension | post |
25 MB STEP cap · 40 MB request cap (413) · per-token rate limit
(MCP_FEA_RATE_LIMIT/_WINDOW_S, 429) · concurrent-solve cap
(MCP_FEA_MAX_CONCURRENT) · daily budget (MCP_FEA_MAX_SPAWNS_PER_DAY) ·
kill switch (MCP_FEA_DISABLE_SPAWN=1). CAD parsing never runs in the
endpoint process (see SECURITY.md).
uv sync
curl -Ls https://micro.mamba.pm/api/micromamba/osx-arm64/latest | tar -xj bin/micromamba
./bin/micromamba create -y -p .tools/ccx-env -c conda-forge calculix=2.23
make test # fast suite (protocol, benchmarks, validation, hardening)
make torture # 21 real-world STEP files through the endpoint
make validation-page
uv run uvicorn --factory src.server.app:create_local_app --port 8000 # local serversrc/pipeline/ geometry, meshing, deck, solver, validation, post, render ·
src/materials/ library · src/server/ protocol, tools, tasks, limits,
worker host, Modal app, Apps UI · src/client/ SDK tasks extension ·
tests/ (incl. torture/) · demo/ · docs/ decisions, validation, upstream notes. License: Apache-2.0.
This server cannot be deployed
Structural FEM for AI agents: frames, plates, buckling, seismic and Eurocode checks, solved locally.
Run FDS, OpenFOAM, CalculiX and other engineering simulations on cloudHPC from your AI assistant.
AI-callable calculators and engineering models with real formulas. No hallucinated math.
CAD in ChatGPT and Claude: editable parts, STEP/STL/drawings, in-chat 3D, automatic checks.
