PrintScale MCP has three well-named, action-verb tools with complete Zod schemas and descriptions. Tool names (printscale_upscale, printscale_job, printscale_credits) clearly convey intent. All parameters have type definitions and descriptions. However, output schemas are not formally documented, responses are text-only without structured field definitions. Error handling returns plain text messages rather than actionable recovery guidance. The upscale tool's description is strong (194 chars), but job and credits descriptions are minimal (58 and 60 chars respectively). No parameter validation constraints (min/max for numeric cm/dpi) are documented in descriptions. Missing: pagination support, per-item error details, and structured output schemas.

• Document output schemas for all three tools. printscale_upscale should return {width, height, credits_charged, tiff_url, png_url, jpg_url}; printscale_job should return {id, status, width, height, error?, outputs?}; printscale_credits should return {balance, plan_name}.
• Expand descriptions for printscale_job and printscale_credits to 100+ chars. Include when to call each tool and what data they reveal. E.g., 'Check the status and download URLs of a completed upscale job. Call this after printscale_upscale returns a job_id, or to poll long-running jobs.'
• Add parameter constraints to descriptions: 'cm: long side in centimeters (10 - 500, default 60)' and 'dpi: output resolution (150, 300, or 600 only; default 300)'. Use enums for dpi in Zod.
• Enhance error messages with recovery guidance. E.g., 'Job timed out after 5 minutes. Try again with a smaller image or different mode.' or 'HTTP 401: Invalid API key. Check PRINTSCALE_API_KEY environment variable.'
• Return structured JSON objects instead of plain text. Use server.tool() to return {type: 'text', text: JSON.stringify(result)} or implement a custom response formatter that preserves field types for LLM parsing.