Hot-reload your MCP servers without restarting your AI coding assistant. A transparent development proxy for the Model Context Protocol that enables seamless server restarts during development.
Single tool 'restart_server' with well-structured schema and clear naming. Tool name follows verb_noun convention (restart_server). Input schema is properly defined with JSON Schema types, descriptions for all parameters, and additionalProperties constraint. However, the tool lacks documentation of output format/structure, and descriptions could be more comprehensive regarding when/why to use the tool and what it returns. The tool serves a meta-purpose (managing the proxy itself) rather than exposing domain functionality, which limits evaluation against production tool patterns.
Restart the MCP server with optional configuration updates
Output schema not documented. Tool description does not specify what fields the response contains, making it harder for LLMs to understand the result structure and plan subsequent actions.
Tool description (63 chars) lacks context about when to use the tool and what it returns. Should include: what the restart accomplishes, typical use cases (development/hot-reload), and structure of success/failure response.
No recovery guidance in error cases. If restart fails, the description does not advise the LLM what to do next (retry with force=true, check logs, verify child command syntax).
Parameter 'childCommand' and 'childArgs' descriptions lack format constraints. Should specify what constitutes valid command syntax, whether absolute paths are required, and examples of valid format.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 15 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 40 | - | v1 |