CEDAR MCP Server

Instructions

Implementation Reference

• src/cedar_mcp/server.py:55-55 (registration)

  Registration and handler for the 'get_template' MCP tool using @mcp.tool() decorator. Fetches template from CEDAR API and applies cleaning.
  @mcp.tool()
• src/cedar_mcp/processing.py:367-452 (helper)

  Helper function clean_template_response that transforms and cleans the raw CEDAR template JSON-LD into a simplified structure using Pydantic models for output schema validation.
  def clean_template_response( template_data: Dict[str, Any], bioportal_api_key: str ) -> Dict[str, Any]: """ Clean and transform the raw CEDAR template JSON-LD to simplified YAML structure. Now supports nested objects, arrays, and template elements. Args: template_data: Raw template data from CEDAR (JSON-LD format) bioportal_api_key: BioPortal API key for fetching controlled term values Returns: Cleaned and transformed template data as dictionary (ready for YAML export) """ # Extract template name, preferring schema:name for correct casing template_name = template_data.get("schema:name", "") if not template_name: # Fallback to title if schema:name is empty title = template_data.get("title", "") template_name = ( title.replace(" template schema", "").replace("template schema", "").strip() ) if not template_name: template_name = "Unnamed Template" # Get field order from UI configuration ui_config = template_data.get("_ui", {}) field_order = ui_config.get("order", []) # Get properties section properties = template_data.get("properties", {}) # Transform fields and elements in the specified order output_children: List[Union[FieldDefinition, ElementDefinition]] = [] # Process fields/elements only in UI order since it covers all template items for item_name in field_order: if item_name in properties: item_data = properties[item_name] if isinstance(item_data, dict): item_type = item_data.get("@type", "") if item_type == "https://schema.metadatacenter.org/core/TemplateField": # It's a simple field field_child = _transform_field( item_name, item_data, bioportal_api_key ) output_children.append(field_child) elif ( item_type == "https://schema.metadatacenter.org/core/TemplateElement" ): # It's a template element (possibly an array) element_child = _transform_element( item_name, item_data, bioportal_api_key ) output_children.append(element_child) elif item_data.get("type") == "array" and "items" in item_data: # It's an array - check what type of items it contains items_type = item_data["items"].get("@type", "") if ( items_type == "https://schema.metadatacenter.org/core/TemplateField" ): # Array of fields - treat as a field with array marker field_child = _transform_field( item_name, item_data, bioportal_api_key ) output_children.append(field_child) elif ( items_type == "https://schema.metadatacenter.org/core/TemplateElement" ): # Array of elements - treat as an element element_child = _transform_element( item_name, item_data, bioportal_api_key ) output_children.append(element_child) # Create output template output_template = SimplifiedTemplate( type="template", name=template_name, children=output_children ) # Convert to dictionary for YAML export return output_template.model_dump(exclude_none=True)
• src/cedar_mcp/model.py:77-87 (schema)

  Pydantic model defining the output schema for cleaned templates, used by clean_template_response.
  class SimplifiedTemplate(BaseModel): """ Represents the complete simplified template structure. """ type: str = Field("template", description="Template type") name: str = Field(..., description="Template name") children: List[Union[FieldDefinition, ElementDefinition]] = Field( ..., description="Template fields and elements" )