Fentaris exposes 5 tools with minimal descriptions and incomplete schemas. Most tools lack parameter descriptions entirely. auth_status and status have empty input objects with no properties documented. release_notes description is vague ('visible to maintainers only'). echo lacks a description for its 'message' parameter. No tool documents output schemas or error recovery paths. Tool names follow verb_noun convention (auth_login, auth_status) but descriptions are too brief (10-30 chars) to guide LLM selection. No tool declares permissions, idempotency, or destructiveness. This is typical of a proxy server that exposes minimal agent-facing tools, but the definitions lack the rigor needed for reliable agentic use.

• Expand tool descriptions to 50-200 characters. Example: 'auth_status' → 'Check which OAuth-protected MCP servers the current user has authorized. Returns a list of server names and their auth status. Call this before auth_login to see what is already authorized.'
• Add descriptions to all parameters. For auth_login's 'server' param: 'Name of the OAuth-protected MCP server to authorize (e.g., github, slack). Call auth_status first to see available servers.'
• Document output schemas for each tool. Example for auth_status: 'Returns {servers: [{name: string, authorized: boolean, expires_at?: ISO8601}]}'
• Add error recovery guidance. Example for auth_login: 'If server not found, call auth_status to list available servers. If login fails, check that the server is properly configured.'
• Declare tool annotations in the MCP tool definition. Mark all tools with readOnlyHint=true since they are read-only. This enables LLMs to reason about safety and retry behavior.

Inferred effective spec: 2025-06-18+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry