Loading ToolBench
MCP server for Photo AI Studio — AI photo generation, editing, and video creation for Claude Desktop, Cursor, Windsurf, and other MCP clients
Photo AI Studio MCP has well-structured tool definitions with complete schemas and descriptions. All 8 tools have clear verb-noun naming (get_account, browse_themes, upload_image, generate_photo, edit_image, create_video, get_prediction, list_predictions). Descriptions are detailed and contextual (avg ~180 chars), explaining WHAT each tool does and WHEN to use it. Input schemas are properly typed with enums for constrained values. However, output schemas are not documented, LLMs cannot predict response structure for downstream chaining. Error handling is basic (generic PhotoAIError/InsufficientCreditsError) without recovery guidance. Parameter descriptions could be more prescriptive about constraints (e.g., max file size for upload_image is mentioned in description but not as a formal constraint). Tool composition is strong, tools chain logically (upload → generate/edit → get_prediction). No security issues detected (API key injected via env, not exposed as parameter).
Browse available photo themes/categories for AI photo generation. Filter by gender or search by tag. Returns theme slugs needed for the generate_photo tool.
Create AI videos. Costs 500 credits (video) or 100 credits (product_holder image). Types: - image_to_video: Animate an image with a prompt (image_url, prompt required). 500 credits. - ugc: Generate UGC video with script and voice (script, voice_id required). 500 credits. - product_holder: Product showcase with person holding product (product_image_url required). 500 credits (video) or 100 credits (image).
Edit a photo using one of 19 AI operations. Credit costs: remove_background (10), replace_background (3), all others (100). Operations: - remove_background: Remove image background (image_url required) - replace_background: Replace background with prompt (image_url, prompt required) - retouch: AI retouch with prompt guidance (image_url, prompt required) - skin: Smooth and enhance skin (image_url required) - hair: Change hairstyle (image_url required, hairstyle or reference_image_url) - makeup: Apply AI makeup (image_url required) - age: Change apparent age (image_url, target_age required) - expression: Change facial expression (image_url, target_expression required) - body_type: Modify body type (image_url, target_body_type required) - eye_color: Change eye color (image_url, target_eye_color required) - color_correction: Auto color correction (image_url required) - custom: Custom AI edit with prompt (image_url, prompt required) - clothes: Change clothes (image_url, clothes_image_url required) - try_on: Virtual clothes try-on (image_url, clothes_image_url required) - product: Product placement (image_url, product_image_url required) - upscale: Enhance resolution (image_url required) - restore: Restore old/damaged photo (image_url required) - face_swap: Swap faces between photos (source_image_url, target_image_url required) - extend: Extend image canvas (image_url required) All image URLs must be CDN URLs from upload_image.
Output schemas not documented. LLMs cannot predict response structure for tools like generate_photo, edit_image, create_video. This breaks downstream tool chaining and forces agents to guess field names.
edit_image tool has 19 operations but conditional parameter requirements are not formally expressed. E.g., 'prompt' is required for replace_background/retouch/custom but optional for others. LLMs may pass invalid combinations.
Error handling returns generic PhotoAIError/InsufficientCreditsError without recovery guidance. 'Error (INVALID_THEME): theme not found' tells agent nothing about next steps. Should suggest 'Call browse_themes() to see available themes.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
Generate AI photos from face images using 150+ themed styles or custom prompts. Costs 100 credits per photo. Requires uploading face images first via upload_image. Use browse_themes to discover available categories.
Get your Photo AI Studio account info: credit balance, total predictions, and membership date. Use this to check available credits before generating photos.
Check the status of an async operation (generation, edit, or video). Returns the current state and result URL when complete.
List past predictions (generations, edits, videos) with pagination and optional state filter.
Upload an image to Photo AI Studio CDN. Returns a CDN URL needed by generate_photo, edit_image, and create_video tools. Accepts a public image URL or base64-encoded image data. Max 10MB.
upload_image accepts both image_url and image_base64 but does not document mutual exclusivity or which is preferred. LLMs may pass both or neither, causing ambiguous failures.
generate_photo and create_video support async mode (wait_for_result=false) returning prediction_id, but polling behavior and timeout expectations are not documented. LLMs may not know when to call get_prediction or how long to wait.