Kadenzo MCP demonstrates solid tool design with 14 well-named, action-verb tools (list_, schedule_, get_, update_, cancel_, upload_, post_). All tools have descriptions (avg ~150 chars, within 10-1024 baseline). Input schemas use Zod with type definitions and descriptions for most parameters. However, several parameters lack descriptions (e.g., 'limit', 'offset' in list_posts), and output schemas are not documented, LLMs cannot predict response structure. Error handling returns basic fail() responses without recovery guidance. Security is sound (API key via env var, no secrets in params). Tool composition is clean: each tool has one responsibility, and outputs include IDs for chaining (e.g., account_id, post id). Naming is consistent and verb-first. Missing: output schema documentation, richer error messages, and parameter descriptions for optional fields.

• Document output schemas for all tools. Example: 'Returns {id: string, status: "pending"|"posted"|"failed", accounts: [{platform: string, username: string, status: string}], ...}'. This enables LLMs to plan downstream calls and extract the right fields.
• Add descriptions to pagination parameters: 'limit: number (1-100, default 25), max results per page' and 'offset: number (default 0), skip this many results'.
• Enhance error messages with recovery guidance. Instead of fail(e), return: 'Error: {message}. Next steps: {retry|ask_user|check_config}. Example: if (res.status === 401) return fail(new Error("API key invalid. Set KADENZO_API_KEY in your MCP client config."))'.
• Add input validation to schedule_post: if (!content && !media_urls) throw new Error('At least one of content or media_urls is required'). Return this error in the response so LLM can self-correct.
• Add validate_only parameter to update_post and cancel_post (or create separate preview_update_post tool) so agents can preview changes before committing. This prevents accidental data loss.