cst-mcp
Turns CST Studio Suite into an AI-controllable electromagnetic design lab, enabling conversational THz metamaterial absorber design, unit-cell automation, Floquet port control, batch solve with checkpoint-resume, and surrogate-model predictions.
README
cst-mcp 🛰️
MCP server for CST Studio Suite — THz metamaterial absorber design, unit-cell automation & Floquet port control
<a id="english"></a>
✨ What is this?
cst-mcp turns CST Studio Suite into an AI-controllable electromagnetic design lab through the
Model Context Protocol. Point Claude/Cursor/DSH at it and drive
full-wave simulations conversationally:
- 🧱 One-shot unit-cell builder — MIM absorber stacks (square ring / patch / circular ring), auto-materializes Floquet ports and sets 1 mode by script (~50× faster per point, no GUI clicks)
- 🎛️ Floquet port suite with the correct 2026 API — mode count / scan angles / polarization,
all scriptable (
SetNumberOfModesConsidered, uppercase"Zmax"…). Most public examples use a renamed legacy API that silently fails on CST 2026 — this server is verified against CST 2026. - ⚡ Absorber quick-design workflow — set params → solve → full 200-pt absorption spectrum + A_peak / in-band A_min / ≥90 % bandwidth in one tool call
- 🔁 Checkpoint-resume batch runner — config fingerprints, PID lock, point-level watchdog that survives solver hangs and COM channel death (survived 2100+ unattended points)
- 📈 Surrogate-model hooks — millisecond spectrum prediction & inverse design TopK when a trained pipeline is provided
- 🩺 Self-diagnosis — CST error-log tail tooling and session restart tools for unattended runs
🚀 Quick start
pip install -r requirements.txt # mcp>=1.2,<2 (2.x removed FastMCP path!)
set PYTHONPATH=C:\PATH\TO\CST\AMD64\python_cst_libraries
python src/cst_absorber_mcp.py # stdio MCP server
Register with your MCP client:
{
"mcpServers": {
"cst-mcp": {
"command": "C:\\CST\\Python\\python.exe",
"args": ["path/to/src/cst_absorber_mcp.py"],
"env": { "PYTHONPATH": "C:\\CST\\AMD64\\python_cst_libraries",
"CST_MCP_WORKDIR": "D:\\cst_mcp_work" }
}
}
}
🧰 Tools (23)
| Family | Tools |
|---|---|
| Health | cst_status cst_log_tail cst_session_restart |
| Project | cst_unitcell_build cst_project_open cst_param_set cst_param_list |
| Floquet ports | cst_floquet_set_modes cst_floquet_set_angles cst_floquet_set_polarization cst_port_modes_read |
| Simulate | cst_solve cst_get_absorption cst_read_sparams_offline cst_export_tree_csv |
| Absorber flow | absorber_quick_design absorber_batch absorber_progress |
| Surrogate | absorber_surrogate_predict absorber_inverse_design |
| Docs | doc_export_step doc_export_structure_png |
(see docstrings inside src/cst_absorber_mcp.py for exact signatures)
📐 Verified reference results (real CST 2026 runs)
| Case | Result |
|---|---|
| Metal-ring narrowband (P65/L50/t8/h12/Rs10 Ω/sq) | A_peak = 0.87 @ 1.08 THz |
| Script-set 1-mode vs default 18 modes | ~10 s/pt vs ~462 s/pt |
| Unattended batch | 2099 points, auto-heal from solver hangs |
📚 docs/
reflect_tables.json— reflected method tables of key COM objects (model3d 387 methods, FDSolver 112, Material 589, Port 75, FloquetPort 25 …) discovered on CST 2026results_api_table.json— offline S-parameter reading API surface
⚠️ Notes
- Windows only (COM automation); CST license required.
- Keep project paths ASCII-only — non-ASCII paths break
project.save()(encoding bug). - Use mcp SDK 1.x; v2.0 dropped the FastMCP import path.
<a id="中文"></a>
中文简介
把 CST Studio Suite 变成 AI 可对话操控的电磁设计实验室:一句话建单元胞(方环/贴片/圆环 MIM)、 脚本化设置 Floquet 端口模数/扫描角/极化(CST2026 正确 API,全网首发实测)、一键求解并返回 200 点吸收谱与指标、断点续跑批量仿真(经 2100+ 点无人值守验证)、可选代理模型毫秒级预测与逆向设计。 安装与配置见上方英文区;工具签名见源码 docstring。
License
MIT — see LICENSE。CST Studio Suite 是 Dassault Systèmes 产品,本项目仅通过官方 Python 接口驱动。
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。