AWS Cost Explorer MCP Server

Instructions

Implementation Reference

• server.py:644-732 (handler)

  The handler function for the 'get_ec2_spend_last_day' tool. Decorated with @mcp.tool() for registration. Queries AWS Cost Explorer API for EC2 compute costs grouped by instance type for the previous day. Processes and prints the breakdown, returns raw API response.
  @mcp.tool() async def get_ec2_spend_last_day(params: EC2Params) -> Dict[str, Any]: """ Retrieve EC2 spend for the last day using standard AWS Cost Explorer API. Returns: Dict[str, Any]: The raw response from the AWS Cost Explorer API, or None if an error occurs. """ print(f"get_ec2_spend_last_day, params={params}") # Initialize the Cost Explorer client ce_client = get_aws_service_boto3_client("ce", params.aws_account_id, params.region) # Calculate the time period - last day end_date = datetime.now().strftime('%Y-%m-%d') start_date = (datetime.now() - timedelta(days=1)).strftime('%Y-%m-%d') try: # Make the API call using get_cost_and_usage (standard API) response = ce_client.get_cost_and_usage( TimePeriod={ 'Start': start_date, 'End': end_date }, Granularity='DAILY', Filter={ 'Dimensions': { 'Key': 'SERVICE', 'Values': [ 'Amazon Elastic Compute Cloud - Compute' ] } }, Metrics=[ 'UnblendedCost', 'UsageQuantity' ], GroupBy=[ { 'Type': 'DIMENSION', 'Key': 'INSTANCE_TYPE' } ] ) # Process and print the results print(f"EC2 Spend from {start_date} to {end_date}:") print("-" * 50) total_cost = 0.0 if 'ResultsByTime' in response and response['ResultsByTime']: time_period_data = response['ResultsByTime'][0] if 'Groups' in time_period_data: for group in time_period_data['Groups']: instance_type = group['Keys'][0] cost = float(group['Metrics']['UnblendedCost']['Amount']) currency = group['Metrics']['UnblendedCost']['Unit'] usage = float(group['Metrics']['UsageQuantity']['Amount']) print(f"Instance Type: {instance_type}") print(f"Cost: {cost:.4f} {currency}") print(f"Usage: {usage:.2f}") print("-" * 30) total_cost += cost # If no instance-level breakdown, show total if not time_period_data.get('Groups'): if 'Total' in time_period_data: total = time_period_data['Total'] cost = float(total['UnblendedCost']['Amount']) currency = total['UnblendedCost']['Unit'] print(f"Total EC2 Cost: {cost:.4f} {currency}") else: print("No EC2 costs found for this period") else: print(f"Total EC2 Cost: {total_cost:.4f} {currency if 'currency' in locals() else 'USD'}") # Check if results are estimated if 'Estimated' in time_period_data: print(f"Note: These results are {'estimated' if time_period_data['Estimated'] else 'final'}") return response except Exception as e: print(f"Error retrieving EC2 cost data: {str(e)}") return None
• server.py:52-67 (schema)

  Pydantic model defining input schema for EC2 tools including days (default 1), region (default us-east-1), and optional aws_account_id.
  class EC2Params(BaseModel): """Parameters for retrieving EC2 Cost Explorer information.""" days: int = Field( default=1, description="Number of days to look back for Bedrock logs", ge=1, le=90 ) region: str = Field( default="us-east-1", description="AWS region to retrieve logs from" ) aws_account_id: Optional[str] = Field( description="AWS account id (if different from the current AWS account) of the account for which to get the cost data", default=None )
• server.py:75-134 (helper)

  Utility function to create boto3 clients for AWS services like 'ce' (Cost Explorer), supporting cross-account access via role assumption.
  def get_aws_service_boto3_client(service: str, aws_account_id: Optional[str], region_name: str, account_b_role_name: Optional[str] = CROSS_ACCOUNT_ROLE_NAME): """ Creates a boto3 client for the specified service in this current AWS account or in a different account if an account id is specified. Args: service (str): AWS service name (e.g., 'logs', 'cloudwatch') region_name (str): AWS region (e.g. 'us-east-1') aws_account_id (str): AWS account ID to access, this is the account in which the role is to be assumed account_b_role_name (str): IAM role name to assume Returns: boto3.client: Service client with assumed role credentials """ try: this_account = boto3.client('sts').get_caller_identity()['Account'] if aws_account_id is not None and this_account != aws_account_id: # the request is for a different account, we need to assume a role in that account print(f"Request is for a different account: {aws_account_id}, current account: {this_account}") # Create STS client sts_client = boto3.client('sts') current_identity = sts_client.get_caller_identity() print(f"Current identity: {current_identity}") # Define the role ARN role_arn = f"arn:aws:iam::{aws_account_id}:role/{account_b_role_name}" print(f"Attempting to assume role: {role_arn}") # Assume the role assumed_role = sts_client.assume_role( RoleArn=role_arn, RoleSessionName="CrossAccountSession" ) # Extract temporary credentials credentials = assumed_role['Credentials'] # Create client with assumed role credentials client = boto3.client( service, region_name=region_name, aws_access_key_id=credentials['AccessKeyId'], aws_secret_access_key=credentials['SecretAccessKey'], aws_session_token=credentials['SessionToken'] ) print(f"Successfully created cross-account client for {service} in account {aws_account_id}") return client else: client = boto3.client( service, region_name=region_name ) print(f"Successfully created client for {service} in the current AWS account {this_account}") return client except Exception as e: print(f"Error creating cross-account client for {service}: {e}") raise e
• server.py:644-644 (registration)

  The @mcp.tool() decorator registers the get_ec2_spend_last_day function as an MCP tool.
  @mcp.tool()