ELF-mcp-server 1.37.0

MCP server for ELF600 electromagnetic field analysis: 31 curated topics (incl. motor workflows and ELF MAGIC -> Cauer Ladder Network synthesis) + public workflow recipes + 332 ELF-runnable public .mai/.meg motor sample decks + Python-interface team28 seed manifest + 1141 help files + 332 examples + 146 wiki pages + Python ctypes API — 24 tools, all searchable.

ELF-mcp-server

MCP server providing ELF600 electromagnetic field analysis documentation — file formats, solver options, element types, workflow recipes, and ELF-runnable public input decks for the ELF600 BEM-based electromagnetic analysis suite (MAGIC magnetostatic, ELFIN electrostatic, BEAM particle tracking).

This server does not execute ELF600 simulations — it provides curated documentation and public .mai/.meg input decks that AI coding assistants (Claude Code, Cursor, etc.) can consult while authoring ELF input files.

---

Features

24 tools + 1 prompt providing curated docs, workflow recipes, ELF-runnable public sample decks, and raw access to ELF600 help HTM, example inputs, vendor wiki, and Python ctypes API:

Tool family	Purpose	Files
elf_usage(topic)	31 curated topics — high-level recipes	(knowledge.py)
elf_help_*(...)	Help HTM files from C:/ELF600/help/	1141 files, 1.18M chars
elf_examples_*(...)	Example .mai/.mei/.txt plus 100-card playbook from C:/ELF600/examples/	332 files, 533k chars
elf_sample_decks_*(...)	Lab-authored ELF-runnable public .mai/.meg sample decks	332 motor cases, 664 input files
elf_recipe_*(...)	Workflow decision cards for elements, PRE/SOL blocks, outputs, checks, and pitfalls	public-safe recipes
elf_wiki_*(...)	Vendor wiki pages from elf.co.jp PukiWiki	146 pages, 211k chars
elf_python_*(...)	Python ctypes API + configs from C:/ELF600/bin/	15 files, 246k chars

Each _* family has 3 tools: _index, _search(query), _get(path). The examples family also has elf_examples_playbook(limit=100), which summarizes 100 .mai examples as compact cards with detected SOL blocks, element families, feature tags, companion .mei/.model files, and reuse hints. The recipe family also has elf_plan_workflow(goal), which chooses a short public-safe recipe sequence from a natural-language analysis goal. The sample deck family has elf_sample_decks_index/search/get/playbook for ELF-runnable public .mai/.meg decks. The Python family also has elf_python_team28(): a compact 28-case seed manifest from the 332 public motor cases for ELF Python-interface orchestration. team28 is not a normal ELF GUI/CLI deck-execution workflow. Solver outputs, comparison metrics, executable orchestration state, and Python-interface runtime state are not bundled.

MCP quick start

For MCP clients, start with elf_overview() to discover the server surface and public boundary. The most useful calls while authoring motor inputs are:

• elf_sample_decks_playbook(limit=20, family="pm_square") for compact cards over the public motor decks
• elf_sample_decks_search("HBCN FLUM", ext="mai") to find reusable input patterns
• elf_sample_decks_get("motor/pm_cosine_pickup_72/pm001/pm001.mai") to open a concrete public deck
• elf_python_team28() to inspect the Python-interface seed manifest

This MCP server is documentation and input-deck retrieval only; it does not launch ELF, run solvers, manage licenses, or publish validation outputs.

ELF/MAGIC for motor input authoring

ELF/MAGIC is useful for motor-oriented magnetostatic input authoring when the model is expressed as .mai analysis control plus .meg mesh data. This server turns that knowledge into MCP tools:

• 332 public motor input-deck pairs covering 2-pole, 4-pole, 6-pole, 8-pole, and cosine-remanence PM pickup families
• playbook cards that expose each deck's SOL blocks, PRE keywords, element families, feature tags, and reuse hints
• curated motor topics for air-gap field, flux linkage/back-EMF pickup, polarity/angle conventions, force outputs, and eddy-current setup
• Python-interface team28 seed manifest for higher-level orchestration, without bundling runtime state or solver outputs

Useful entry points are elf_usage(topic="ipm_motor"), elf_usage(topic="motor_radia_bridge"), elf_recipe_search("motor pickup"), and elf_sample_decks_playbook(limit=50, family="pm_square").

Public lint

Before publishing, run:

python -m pytest
elf-mcp-server --selftest
elf-mcp-policy-lint

elf-mcp-policy-lint checks the public package boundary: no private validation paths, no unrelated commercial-tool references, and no bundled solver outputs inside the public motor sample decks.

Bundled data (all generated from fresh ELF600 install via scripts/crawl_*.py):

• help_dump.json — Shift_JIS HTM decoded + HTML-stripped
• examples_dump.json — 228 MAGIC + 66 ELFIN + 38 BEAM input files
• wiki_dump.json — 146 curated pages from https://elf.co.jp/
• python_dump.json — elftypes.py/magtypes.py (83 ctypes API functions each), *.cfg, ELFERR.def/MESERR.def, etc.

Curated topics (elf_usage)

Returns documentation on:

• File formats: .mai (analysis input), .mei (mesh script), .meg (compiled mesh)
• Solvers: MAGIC (magnetostatic, transient, AC), ELFIN (electrostatic), BEAM (particle tracking)
• Eddy current: MAB / MAT / MBB elements, time-stepping, sinusoidal AC (SOL MOMC)
• Element types: full catalog with DOF counts and symmetry restrictions (3D / 2D / Axisym)
• B-H curves: anisotropy (HBA1/HBA2), recoil, extrapolation
• Motor workflows: IPM Ld/Lq plus a radia-mcp/open-FEA concept bridge for air-gap field, torque, flux linkage/back-EMF, lamination, and eddy-current studies
• Inductance: Lsc (JIS) and Ll (IEEJ) with 6 samples
• Magnetization / demagnetization (MAGNE2)
• Convergence troubleshooting, error codes (160+ ELF-Q/E/W codes)
• Force methods: FORC vs FORT vs FIXB
• Tools: IEmesh, Wmap3, MagFilter2, MaiEditor3, ELF/Bench

Available topics:

all, overview, mai_format, mei_format, meg_format,
magic, elfin, beam, element_types, bh_curves,
sol_commands, mei_commands, ipm_motor, motor_radia_bridge, inductance,
magnetization, examples, meg_export, treasure_box,
sinusoidal, anisotropy, sted, meshing, convergence,
force_methods, errors, iemesh, tools, cln_extraction,
licensing, python_api, live_drive

The cln_extraction topic documents the 6-step ELF MAGIC -> Cauer Ladder Network synthesis workflow (Foster fit + Cauer-I/II + 3-way validation against step response / Joule loss / Lorentz force). Distilled from a 21-script rectangular-CLN reference analysis suite.

---

Installation

pip install ELF-mcp-server

The current PyPI distribution name is ELF-mcp-server. The installed console scripts remain lowercase: elf-mcp-server and elf-mcp-policy-lint.

Verify:

elf-mcp-server --selftest

---

Usage

Claude Code

claude mcp add elf "C:/Program Files/Python312/Scripts/elf-mcp-server.exe"

(Adjust path for your Python install. On Linux/macOS, the script is typically ~/.local/bin/elf-mcp-server or similar.)

Cursor / Other MCP clients

Add to your MCP config:

{
  "mcpServers": {
    "elf": {
      "command": "elf-mcp-server"
    }
  }
}

Self-test

elf-mcp-server --selftest

Iterates through all curated topics and asserts non-empty documentation.

---

What is ELF600?

ELF600 is a commercial BEM (Boundary Element Method) electromagnetic analysis suite distributed by Science Solutions International Laboratory (https://www.science-solutions.jp/elf/).

Module	Purpose
MAGIC	Magnetostatic field (static, transient, AC). Eddy current via MAB/MAT/MBB.
ELFIN	Electrostatic field analysis (D-E curves)
BEAM	Charged particle beam tracking

Workflow:

.mei (mesh script) --> IEmesh --> .meg (compiled mesh)
.mai (analysis) + .meg --> MAGIC/ELFIN/BEAM --> .mag/.mao results

---

Why this server?

LLM coding agents authoring ELF input files (.mai/.mei) need access to:

• The ~60k character ELF reference manual content,
• Element naming conventions (T/K/R symmetry × element family),
• SOL block recipes (MOME / MOMC / FIEL / FORC / NONL),
• Frequency-sweep AC analysis structure,
• Common error code interpretation,

without polluting context with the entire vendor PDF. This MCP server returns just the relevant topical chunk on demand.

---

License

BSD-3-Clause. See LICENSE.

ELF600 itself is a commercial product of Science Solutions International Laboratory and is not redistributed by this package — only documentation references.

---

Author

Kengo Sugahara, Kindai University ([e-mail])