MCP server for generating images using NovelAI API. Supports character-prompts.md parsing for automatic character illustration generation, generic image generation, cover generation, and multi-character composition.
The server provides 6 tools with basic input schemas and descriptions, but suffers from significant gaps in definition quality. Tool descriptions are present but relatively generic (averaging 72-98 characters, below the 194-char baseline). Parameter descriptions exist but lack precision around constraints, formats, and ranges. No output schemas are documented, critical for agent planning. Error handling is minimal and doesn't guide recovery. The tools show reasonable naming (verb_noun pattern), but parameter type coverage is incomplete in places, and no tool annotations (readOnlyHint/destructiveHint) are present. This server falls in the 'poor' range (50-59 adjusted downward for documentation gaps).
Generate character image from character-prompts.md with optional variant selection
Generate a NovelAI image with custom prompt and parameters
Generate image with multiple characters at specified positions (composition)
Get the prompt details for a specific character and variant
Get style configuration (prefix and negative tags) from a novel's character-prompts.md
List all available characters from a novel's character-prompts.md file
No documented output schemas. Tools return image bytes, objects, or lists, but LLMs cannot plan downstream calls or extract required fields without knowing the response structure.
Missing tool annotations. No readOnlyHint/destructiveHint/idempotentHint declared. Agents cannot infer which tools are safe to retry (critical for error recovery).
Insufficient parameter constraint documentation. Numeric parameters (width, height, steps, scale, seed) lack explicit min/max bounds in descriptions. 'default 832, must be multiple of 64' appears only in one tool; consistency missing.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 40 | 2026-07-28+ | v2 |
Generic descriptions lack WHEN/WHY context. 'Generate a NovelAI image...' explains WHAT, not WHEN to use generate_image vs. generate_character_image vs. generate_multi_character_image. LLMs cannot disambiguate.
No error recovery guidance. Code references NovelAI API errors (401 auth, 402 Anlas quota, 429 rate limit, 500 internal), but tool descriptions do not document these failure modes or what the agent should do (retry, adjust parameters, ask user).
generate_multi_character_image 'characters' parameter has complex nested schema (array of objects with name, variant, x, y) but is not validated or documented for required vs. optional fields. Missing description of position ranges, valid variant keys, etc.
Missing dependency hints. No guidance that generate_character_image and generate_multi_character_image require a valid novel_id (which must first be discovered via list_characters or other means).
list_characters and get_character_prompt return unspecified structures. No documentation of whether results include variant names, prompt text, style overrides, etc.
list_characters has no pagination. If a novel has many characters, response could explode; no limit, offset, or pagination guidance.
seed parameter accepts 0 for random, but other numeric constraints (width, height, steps) use defaults without explaining randomization behavior or idempotency.