MCP server for PromptPilot.club — generate images, video, and audio via Pollinations API
PromptPilot MCP Server demonstrates solid foundational quality with well-structured tool definitions and comprehensive schemas. All 8 tools have descriptions, Zod schemas with typed parameters, and clear naming conventions. However, several definition quality patterns are inconsistently applied: descriptions lack actionable depth for agent selection, output schemas are not explicitly documented, and error handling guidance is minimal. The server implements a reasonable composition of related tools (list_models, build_prompt, generate_*), but lacks recovery hints and actionable error messages. Per-tool average: 68/100.
Build an optimized prompt from subject + style/lighting/camera/mood/color/quality options
Check your Pollinations API balance (pollen). Shows remaining credits for paid models.
Generate audio (speech or music) using Pollinations API. Returns a URL to the generated audio.
Generate multiple images in one call (up to 10). More efficient than calling generate_image repeatedly.
Generate an image using Pollinations API. Returns a URL to the generated image.
Generate a video using Pollinations API. Returns a URL to the generated video.
Output schemas not documented. Users cannot introspect what fields list_models, generate_image, or generate_batch return. LLMs must infer structure from text responses, risking parsing errors.
Descriptions lack actionable context for LLM selection. 'Generate an image using Pollinations API' does not tell the LLM WHEN to use generate_image vs generate_batch, or what to do if the Pollinations API is unavailable. Should include use-case hints and error recovery paths.
Error handling lacks recovery guidance. handleGenerateImage returns isError=true with 'Unknown image model' but does not guide the LLM to call list_models. Tool should return: 'Unknown image model: {model}. Call list_models to see available models (flux, dall-e, etc.)'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 60 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 50 | - | v1 |
List all available models for image, video, and audio generation with their type and pricing
List all available styles, lighting, camera angles, moods, colors, and quality tags for prompt building
API key injection as environment variable is secure, but no mention in tool descriptions that paid models require POLLINATIONS_API_KEY setup. Users/agents trying to use 'dall-e' or other paid models will fail silently without guidance.
Parameter 'enhance' (boolean, defaults true) in generate_image lacks description of what 'enhance' means. Does it add keywords? Rephrase grammar? Change artistic style? The description must explain the side effect explicitly.
generate_batch accepts up to 10 prompts but no documentation of response structure (per-item success/failure, partial completion handling). If 1 of 10 fails, does the tool return partial results or full error?
Tool composition opportunity: build_prompt + generate_image are adjacent but neither tool description hints at the pipeline. LLMs may not automatically chain them. Consider: 'Common pattern: call build_prompt first to optimize your subject/style/lighting, then pass result to generate_image.'
No pagination, limits, or result count guidance. check_balance description does not state what format the balance is returned in (integer pollen credits? decimal? currency symbol?). generate_batch silently caps at 10 but does not return a 'capped' indicator.