MCP Salesforce Connector

Instructions

Implementation Reference

• src/salesforce/server.py:56-82 (handler)

  Core handler function in SalesforceClient that fetches object fields using describe(), filters relevant info (label, name, updateable, type, length, picklistValues), caches results, and returns formatted JSON.

  def get_object_fields(self, object_name: str) -> str:
      """Retrieves field Names, labels and typesfor a specific Salesforce object.
  
      Args:
          object_name (str): The name of the Salesforce object.
  
      Returns:
          str: JSON representation of the object fields.
      """
      if not self.sf:
          raise ValueError("Salesforce connection not established.")
      if object_name not in self.sobjects_cache:
          sf_object = getattr(self.sf, object_name)
          fields = sf_object.describe()['fields']
          filtered_fields = []
          for field in fields:
              filtered_fields.append({
                  'label': field['label'],
                  'name': field['name'],
                  'updateable': field['updateable'],
                  'type': field['type'],
                  'length': field['length'],
                  'picklistValues': field['picklistValues']
              })
          self.sobjects_cache[object_name] = filtered_fields
          
      return json.dumps(self.sobjects_cache[object_name], indent=2)

• src/salesforce/server.py:133-146 (registration)

  Tool registration in list_tools() defining name, description, and input schema.

  types.Tool(
      name="get_object_fields",
      description="Retrieves field Names, labels and types for a specific Salesforce object",
      inputSchema={
          "type": "object",
          "properties": {
              "object_name": {
                  "type": "string",
                  "description": "The name of the Salesforce object (e.g., 'Account', 'Contact')",
              },
          },
          "required": ["object_name"],
      },
  ),

• src/salesforce/server.py:136-145 (schema)

  JSON Schema for tool input validation: requires 'object_name' string.

  inputSchema={
      "type": "object",
      "properties": {
          "object_name": {
              "type": "string",
              "description": "The name of the Salesforce object (e.g., 'Account', 'Contact')",
          },
      },
      "required": ["object_name"],
  },

• src/salesforce/server.py:339-351 (handler)

  MCP server tool dispatcher handler that extracts arguments, calls the core get_object_fields, and formats response as TextContent.

  elif name == "get_object_fields":
      object_name = arguments.get("object_name")
      if not object_name:
          raise ValueError("Missing 'object_name' argument")
      if not sf_client.sf:
          raise ValueError("Salesforce connection not established.")
      results = sf_client.get_object_fields(object_name)
      return [
          types.TextContent(
              type="text",
              text=f"{object_name} Metadata (JSON):\n{results}",
          )
      ]