A comprehensive MCP server providing AWS CLI operations, storage management, CloudWatch logs, code execution, Wikipedia search, and general utilities through multiple MCP tool servers
This MCP server has significant structural and quality issues. Of 28 tools, almost none have complete, LLM-optimized descriptions. Parameter schemas are present but descriptions are often generic or missing entirely. Many tools lack clear purpose differentiation (e.g., multiple S3 listing tools with overlapping functionality). Error handling is minimal, most tools just return stdout/stderr without guidance. The codebase shows basic Python MCP integration but lacks production-grade tool design patterns. Only 6 of 28 tools have descriptions exceeding 50 characters; most are under 30 chars, violating baseline description standards. No tool demonstrates recovery guidance, permission gating, or structured error responses. The naming convention is inconsistent (run_aws_cli vs list_buckets vs get_ec2_instances). No visible input validation or constraint documentation in schema. This server would not pass code review for production agent use.
Check if a bucket is empty by listing its objects.
Create a new secret in AWS Secrets Manager.
Search book list by keyword and then return book list
Returns the current date and time in the specified format
Get EBS volumes usage information
Retrieve EC2 instance information using AWS CLI.
Get EFS file systems usage information using CloudWatch metrics
Severe description deficiency: 18 of 28 tools have descriptions under 50 characters, and 8 tools have descriptions under 30 characters. Rubric baseline requires 10 - 1024 chars with optimal 50-200 chars for LLM optimization. Tools like 'list_groups', 'get_logs', 'get_book_list', 'search', 'summary', and 'page' provide minimal context for LLM tool selection.
Duplicate and overlapping tool functionality with no clear LLM guidance on which to choose: list_objects_by_cli, list_objects, and list_buckets all perform S3 object listing with identical semantics. Similarly, list_s3_buckets and list_buckets both list buckets. This violates pattern:tool (one clear job per tool) and creates reasoning overhead.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 9 | - | v1 |
Get CloudWatch logs from a specific log group and stream.
Calculate total storage usage across all S3 buckets
retrieve weather information by city name and then return weather statement.
List S3 buckets using async client with pagination
List available CloudWatch log groups.
List objects in a specific bucket using async client with pagination
List objects in an S3 bucket using list-objects-v2 command.
List S3 buckets and their contents as resources with pagination
List all S3 buckets in the specified region.
List secrets in AWS Secrets Manager.
Get a full Wikipedia page
Get a random Wikipedia page
Use this to execute python code and do math. If you want to see the output of a value, you should print it out with `print(...)`. This is visible to the user.
Execute a Python script for draw a graph. Since Python runtime cannot use external APIs, necessary data must be included in the code. The graph should use English exclusively for all textual elements. Do not save pictures locally bacause the runtime does not have filesystem. When a comparison is made, all arrays must be of the same length.
Retrieve AWS costs using AWS Cost Explorer for the specified period.
run aws command using aws cli and then return the result
Search Wikipedia for a query
Set the Wikipedia language
Retrieve accurate stock data for a given ticker.
Get summary of a Wikipedia page
Execute AWS service operations using boto3 with comprehensive error handling and validation. This tool provides a universal interface to AWS services, allowing you to execute any operation supported by boto3.
No input validation or actionable error handling. Source code shows tools return raw subprocess stderr (e.g., mcp_server_aws_cli.py line ~45: 'error_message = f"execution error: {result.stderr}"'). Per pattern:recovery-guide, errors must tell the LLM what to do next. A 'Bucket does not exist' error should suggest 'try list_s3_buckets()' or 'check credentials'. Current errors give no recovery path.
Missing parameter descriptions or trivial descriptions for key inputs: run_aws_cli parameters (command, subcommand, options) lack guidance on format constraints. retrieve_aws_cost's 'granularity' param has no enum values documented. list_objects_by_cli's 'prefix' and 'delimiter' params lack examples of valid syntax. repl_coder and repl_drawer both expose 'code' param with generic description but no constraint documentation (timeout? max size? safe operations only?).
No output schema documentation. Tools return raw strings (subprocess output or API responses) without documenting field structure. E.g., get_ec2_instances returns JSON from AWS CLI but no schema documenting instance fields, tags, or how to chain into subsequent tools. Pattern:tool requires documented return types.
No idempotency guarantees or side-effect warnings. create_secret, repl_coder, and use_aws are destructive/write operations but descriptions do not warn that repeated calls create duplicates or modify state. Per pattern:command-tool, write operations must be clearly marked and document retry safety.
No permission gating or scope declaration. use_aws tool accepts arbitrary AWS service/operation names with minimal validation. Per pattern:permission-gate and pattern:scope-declaration, destructive operations should verify caller permissions and declare required scopes. No audit trail visible.
Inconsistent naming convention across tools. Names like 'run_aws_cli', 'get_logs', 'repl_coder', 'stock_data_lookup', and 'use_aws' do not follow consistent verb_noun patterns. 'run_aws_cli' is redundant; should be 'execute_aws_command'. 'repl_coder' and 'repl_drawer' are vague about what they control.
No pagination or result limits documented. list_objects, list_buckets, get_logs, list_secrets, and get_ec2_instances can return unbounded results. Per pattern:paginated-result, tools returning lists must accept limit/offset and document max result count. Risk of context window exhaustion.
No tool annotation hints (readOnlyHint, destructiveHint, idempotentHint). Per protocol spec baseline, tools should declare their impact category. This server registers 0 tool annotations despite having ~12 write/destructive operations (create_secret, use_aws, repl_coder, repl_drawer, set_lang, etc.).