A hello world MCP server using FastMCP with HTTP transport that demonstrates basic MCP functionality including greeting tools, resource serving, and HTTP streaming transport
Server provides 2 tools with reasonable structure but exhibits significant quality gaps. Both tools have descriptions and input schemas present, but parameter descriptions are minimal, output schemas lack documentation, and error handling is absent. The say_hello tool has acceptable parameter documentation with type and default values clearly specified in the schema; get_server_info has no input parameters. Neither tool provides guidance for error recovery or documents expected output structure. Naming follows verb-noun convention (say_hello, get_server_info) appropriately. Tool definitions are explicitly registered via @mcp.tool() decorators in mcp_hello/server.py, so all tools are directly visible in source. Average per-tool score: 58.
Get information about this MCP server.
A simple greeting tool that says hello in different languages.
Output schemas not documented for any tool. say_hello returns {'greeting', 'language', 'name', 'message'} and get_server_info returns {'name', 'version', 'description', 'capabilities', 'supported_languages'}, but these structures are inferred from implementation only, no JSON Schema definitions provided to the LLM. LLMs cannot predict downstream field availability for chaining tools.
Error handling completely absent. No tool provides guidance when invalid language code is passed to say_hello (e.g. 'zz' not in greetings dict, silently falls back to 'en' without informing LLM). No error categorization (retryable, user-fixable, fatal). LLMs cannot diagnose failures or adapt strategy.
Parameter descriptions insufficient for LLM reasoning. 'language' parameter description 'The language code for the greeting (en, es, fr, de, it, pt, ru, ja, ko, zh)' lists values but does not explain what happens on unsupported input or state a constraint. 'name' has 'The name to greet', generic and uninformative about character limits, special characters, or length bounds.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 46 | - | v1 |
get_server_info has overly brief description ('Get information about this MCP server.'), 34 chars, below the 50-char baseline for adequate context. Does not explain WHEN to call this vs other discovery methods or what decision the LLM should make with the returned info. Baseline description length for A+ tools: 194 chars (p10=34, p90=392). This is at the p10 floor.
Missing output schema documentation. Resources (file://hello-world, file://server-status) are implemented but tool definitions lack descriptions of return types. Rubric baseline: 100% of A+ tools have documented return types. Current: 0%.