MCP server for Yandex.Direct API integration, providing tools for managing advertising campaigns, keywords, ads, and account settings
Server has 3 tools with significant definition gaps. auth_status and auth_setup have basic schemas and descriptions, but descriptions are vague and lack actionable context. tool_help is a meta-tool with minimal utility for end users. No parameter descriptions for auth_setup's 'login' field. Schemas present but descriptions fail to explain WHEN to use each tool, WHAT happens on success, or recovery paths on failure. No output schemas documented. No error handling guidance. Naming is acceptable (verb_noun pattern) but descriptions are too generic to guide LLM tool selection effectively.
Save a ready-made direct OAuth token (y0_...) into a direct auth profile; use auth_login for the interactive browser flow. Call tool_help('auth_setup') for parameters.
Check the current direct auth profile status (token presence, login, expiry). Call tool_help('auth_status') for parameters.
Get the full documentation for another tool: parameter reference, examples and constraints. Call this BEFORE using an unfamiliar tool — the tools themselves carry only a one-line summary. Pass the exact tool name (e.g. 'campaigns_add'); omit the name to list every available tool.
auth_setup description lacks clarity on prerequisites and side effects. Does not state that this WRITES credentials to disk, making it a destructive operation requiring confirmation. LLM cannot determine if this is safe to retry.
auth_setup 'login' parameter has no description. LLM cannot determine what format is expected (email, username, numeric ID?) or whether it is required.
No output schemas documented for any tool. LLM cannot plan downstream calls or extract required fields. auth_status returns a dict with 'valid', 'profile', 'expires_in_human' but this is not declared in tool definition.
auth_status description delegates to tool_help('auth_status') for parameter docs. This forces an extra tool call before the LLM can use the tool, violates single-call discoverability. Parameters should be self-documenting.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
No error handling guidance. If auth_setup fails (invalid token, network error, permission denied), LLM has no recovery path. Descriptions do not explain what errors are retryable vs user-fixable.