Configuration toolchain for the MCP Gateway: generate starter configs from curated deployment shapes, validate against the gateway's real schema, explain effective wiring, scan for inline secrets, and emit the JSON schema.
mcpg-config is a CLI tool wrapped as MCP with 7 subcommands. Tool definitions are inferred from clap argument parsing rather than explicit MCP tool registration with schemas. No input schemas are visible in the provided source, only clap derive macros and argument descriptions. Descriptions are present but brief (10-50 chars). Parameters lack formal type constraints and descriptions. The tool is domain-specific (config validation) but lacks the structured output documentation and error guidance expected of production MCP tools. Transport is unknown (likely STDIO via clap).
Validate a config (deny_unknown_fields) before booting
Render the configuration reference from the live schema
Explain a config field / resolve env + cred references
Scaffold a validated starter config from a template
Emit the JSON Schema for the AppConfig
Inspect cred:// + ${env} references
Show how bindings wire to backends + plugins
No input schemas visible. Tool definitions inferred from clap derive macros, not explicit MCP tool registration. Cannot verify parameter types, constraints, or structured output schemas.
Tool descriptions are too brief (10-50 chars). Lack context on WHEN to use each tool, WHAT it returns, and prerequisites. E.g., 'Validate a config' does not explain what validation rules apply or what happens on failure.
Parameter descriptions missing or minimal. 'template' param in init has a description, but 'force' and 'list' lack clarity on mutual exclusivity and side effects. No enum constraints visible for template names.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
No output schema documentation. Tools like 'doc' and 'schema' emit Markdown and JSON respectively, but no structured response schema is defined for LLM consumption. Agents cannot plan downstream calls.
Error handling not documented. No guidance on retryability, user-fixable vs fatal errors, or recovery steps. E.g., what happens if a config file is malformed? What should the agent do next?