An MCP server that provides tools to interact with AWS Cost Explorer API, retrieve Bedrock invocation logs, and analyze AWS cloud spending
AWS Cost Explorer MCP server has 4 tools with moderate definition quality. Tool naming follows verb_noun convention (get_*), which is correct. However, parameter descriptions have issues: several lack specificity (e.g., 'Number of days to look back for Bedrock logs' appears verbatim in EC2Params despite being for EC2, not Bedrock), parameter constraints are partially present but inconsistently applied, and output schemas are not documented anywhere in the provided source. Tool descriptions are present but generic (34-82 chars), falling below the 50-200 char A+ baseline. The 'system_prompt_for_agent' tool is unusual and poorly specified, it takes only an aws_account_id and returns a prompt string, with no schema documentation. Error handling is absent from visible code. The schema quality is hindered by incomplete documentation of return types across all tools.
Retrieve Bedrock daily usage statistics for the last n days in a given region
Retrieve Bedrock invocation logs for the last n days in a given region as a dataframe
Retrieve EC2 Cost Explorer information for the last n days in a given region
Generates a system prompt for an AWS cost analysis agent
Missing output schemas for all tools. No documentation of what fields each tool returns (e.g., get_bedrock_logs returns a DataFrame with specific columns, but this is only visible in code comments, not in the tool definition visible to LLMs). Tools should declare return type, fields, and structure in the schema.
Copy-paste error in EC2Params: 'days' parameter description says 'Number of days to look back for Bedrock logs' but should say 'EC2 costs' or similar. This misleads LLMs about the tool's purpose.
Tool descriptions are too short (34-82 chars). Baseline for A+ tools is 50-200 chars; this server's descriptions (e.g., 'Retrieve Bedrock daily usage statistics for the last n days in a given region') lack context about WHEN to use each tool vs. similar tools or WHY an agent would call it.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 56 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 0 | - | v1 |
'system_prompt_for_agent' is poorly specified. It takes an aws_account_id parameter and returns a prompt string. The purpose is unclear to an LLM, is this a discovery tool? A helper for other agents? The description says 'Generates a system prompt' but does not explain what the LLM should do with it or when to call it.
No error handling documentation. None of the tools describe what happens on failure, what exceptions might occur, or how an LLM should recover. E.g., if logs cannot be fetched, should the LLM retry, ask the user for a different date range, or check permissions?
Parameter 'aws_account_id' is used to support cross-account role assumption (via environment variable CROSS_ACCOUNT_ROLE_NAME). This design couples the tool to an environment variable not under user control. The description does not mention the prerequisite (that the role must exist and be assumable).