metamodel
AI-callable calculators and engineering models with real formulas. No hallucinated math.
我该使用它吗
质量与安全性
基于对工具定义和协议合规性的自动分析。
上下文开销
这是每次将服务器的工具加载到模型上下文窗口时所消耗的大致 token 数。数值越高，可用于其他任务的注意力就越少。
安装
一键安装
将以下内容添加到你的 `claude_desktop_config.json` 文件中：
{
"mcpServers": {
"metamodel": {
"command": "npx",
"args": [
"metamodel-mcp"
]
}
}
}可运行的软件包
1.1.2stdio远程端点
https://www.metamodel.app/api/mcpstreamable-http它能做什么
工具清单
工具（4）
🟢metamodel_list_projects
Browse available MetaModel calculators and engineering models. Returns project names, descriptions, tokens, and model names for each published project.
输入模式
{
"type": "object",
"properties": {},
"additionalProperties": false
}🟢metamodel_get_schema(token)
Get the input/output schema for a published MetaModel project. Returns model names, input parameters (with types, defaults, validation), and output fields. Models marked "presentation": true are display scaffolding (3D scenes etc.) — skip them for engineering analysis. Use a token from metamodel_list_projects.
输入模式
{
"type": "object",
"properties": {
"token": {
"type": "string",
"description": "The publish token of the project (from metamodel_list_projects)"
}
},
"required": [
"token"
],
"additionalProperties": false
}⚪metamodel_export_kn(token)
Export a published MetaModel project as KN notation — a human-readable text spelling of the model: nested Component blocks, "Name: value" literals, "Name: expression" formulas, dotted cross-references. Deliberately lossy (rules, indicators, validations, table data and views are omitted). Useful for reading or diffing the engineering knowledge behind a project.
输入模式
{
"type": "object",
"properties": {
"token": {
"type": "string",
"description": "The publish token of the project (from metamodel_list_projects)"
}
},
"required": [
"token"
]
}🟢metamodel_compute(token, model, inputs)
Run a computation on a MetaModel project. Send input values and get computed outputs from all models. Inputs are auto-routed to the correct model by property name. Models marked "presentation": true in results are display scaffolding, not engineering outputs. Use metamodel_get_schema first to discover available inputs and outputs.
输入模式
{
"type": "object",
"properties": {
"token": {
"type": "string",
"description": "The publish token of the project"
},
"model": {
"type": "string",
"description": "Target a specific model by name. When omitted, returns outputs from ALL models (recommended)."
},
"inputs": {
"type": "object",
"description": "Input values as key-value pairs (e.g., {\"width\": 12}). Auto-routed to the correct model. Omitted inputs use defaults.",
"additionalProperties": true
}
},
"required": [
"token"
],
"additionalProperties": false
}社区
证据