MCP server for interacting with Avito API. Provides tools for OAuth2 token management and OpenAPI endpoint lookup/documentation.
Four tools with complete schemas and descriptions, but significant gaps in parameter documentation and error handling guidance. Tool names follow verb_noun convention (get_token, search_endpoints, show_endpoint, list_tags). Descriptions are present and substantive (100-200 chars), meeting baseline. However, parameter descriptions are sparse or missing context. No output schemas documented. Error handling is minimal, no recovery guidance, no categorization of retryable vs fatal errors. The server exposes OAuth credentials as parameters (client_id, client_secret), violating secret-injection pattern. No pagination support documented despite search_endpoints potentially returning large result sets.
Get an Avito OAuth2 access token via Client Credentials flow. Requires AVITO_CLIENT_ID and AVITO_CLIENT_SECRET environment variables. Token TTL is 24h; cache it yourself if you make many calls. Avito returns 401 for expired tokens — refresh and retry.
List all Avito API endpoint categories (tags) with endpoint counts. Returns tags sorted by count (descending) and alphabetically.
Search for Avito API endpoints by keyword in path/summary/tag (case-insensitive). Supports Russian word stem matching for inflections. Returns matching endpoints with method, path, summary, and tags.
Show full details of a specific Avito API endpoint including operation object, parameters, request/response schemas (with $ref resolved one level), and rate limits. Supports fuzzy path matching.
OAuth credentials (client_id, client_secret) exposed as tool parameters in get_token. Credentials must never appear in parameters, use server-side secret injection via environment variables. Agent traces log all parameters, leaking secrets into logs and prompt history.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure. search_endpoints and show_endpoint return JSON but structure is undocumented.
Parameter descriptions lack actionable constraints. 'query' in search_endpoints says 'optional, case-insensitive' but does not explain what happens if empty, max length, or expected format. 'exact' boolean lacks explanation of what stem-matching does.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
No error handling guidance. Tools can fail (HTTP 401 for expired token, path not found in show_endpoint) but descriptions do not tell LLM what to do next. No recovery hints, no categorization of retryable vs fatal errors.
search_endpoints can return up to 50 results but no pagination support documented. Large result sets waste tokens and degrade LLM reasoning. No limit parameter, no next_cursor, no total count in response.