MCP Server Starter Kit - Minimal Model Context Protocol server template
Single tool 'echo' with minimal but complete schema definition. Tool is properly registered with McpServer.tool() and includes input schema via Zod, description, and async handler. However, the tool is a trivial demonstration with no production value, minimal description depth, and no parameters beyond a single string. The descriptions are present but lack the guidance and context expected for production tools. No error handling, no output schema documentation, no security considerations. This is a starter template, not a production-grade server.
Echo back the provided message. A simple example to demonstrate the MCP tool pattern.
Tool name 'echo' does not follow verb_noun convention. Should be 'echo_message' or similar to clarify the action and match LLM expectations for tool selection. Generic single-word names reduce clarity when multiple tools are available.
Tool description is 118 characters and lacks actionable guidance. Does not state WHAT the tool does (return value format), WHEN to use it, or any prerequisites. 'Echo back the provided message' tells LLM it's a mirror function, but no context for when an agent should invoke it.
Parameter 'message' has type string and description 'The message to echo back', but no constraints (length limits, format). Unbounded string parameters allow LLMs to pass arbitrarily large payloads.
No documented output schema. Tool returns {content: [{type: 'text', text: '...'}]} but LLMs are not told what fields to expect or how to parse the response.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 48 | - | v1 |
No error handling or recovery guidance. If the input is invalid (e.g., extremely large string causing OOM), no actionable error is returned to guide the LLM's next step.