This server is an MCP interface for anime image generation with Animagine XL 4.0, letting AI agents validate, optimize, explain, and generate images.
validate_prompt: Check prompts against Animagine XL rules (quality tags, ordering, tag count, character/series consistency, resolution compatibility).
optimize_prompt: Reorder and complete prompts into canonical tag order from a description or existing prompt.
explain_prompt: Break down prompt tags into categories and explain what each tag affects.
list_models: Discover available checkpoints and LoRAs, plus the default checkpoint and loaded state.
load_checkpoint: Pre-load a checkpoint into GPU memory for faster subsequent generation.
unload_loras: Remove all LoRA weights from the current pipeline without reloading the model.
generate_image: Text-to-image generation with custom checkpoints, LoRAs, size, steps, CFG, seed, negative prompt, and render-type enforcement.
generate_image_from_image: Image-to-image transformation with strength control for style transfer, refinement, composition preservation, and character consistency.
Allows fetching and using models (e.g., Animagine XL 4.0) from Hugging Face for image generation.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Animagine MCP Optimize this prompt for Animagine XL: 1girl, solo, sitting in a cafe"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server for anime image generation with Animagine XL 4.0. Gives AI agents four tools: validate and optimize Danbooru-style prompts, list local models, and generate images (text-to-image or image-to-image) with custom checkpoints and LoRAs.
Python 3.11+
NVIDIA GPU with ~7 GB VRAM recommended (CPU works, but is very slow)
~7 GB of disk for the base model (downloaded from HuggingFace on first generation)
Related MCP server: FastMCP
git clone https://github.com/gabrielalmir/anime-diffusion-mcp.git
cd anime-diffusion-mcp
python -m venv .venv
# Windows: .venv\Scripts\activate | Linux/macOS: source .venv/bin/activate
pip install -e .For CUDA, install the matching PyTorch build first (see https://pytorch.org/get-started/locally/), then pip install -e ..
Add the server to your MCP client (Claude Desktop, Claude Code, Cursor, ...). See .mcp.json.example:
{
"mcpServers": {
"anime-diffusion": {
"command": "anime-diffusion-mcp",
"env": { "CUDA_VISIBLE_DEVICES": "0" }
}
}
}If the client doesn't inherit your virtualenv, point command at the venv script, e.g. C:/path/to/.venv/Scripts/anime-diffusion-mcp.exe.
The server runs over stdio and uses the current working directory for checkpoints/, loras/ and outputs/, so launch it from the project folder (or set cwd in the client config).
Tool | Purpose |
| Check a prompt against Animagine XL rules: quality tags present and last, 8+ tags, character/series consistency, resolution risk. Returns |
| Reorder tags into canonical order (subject → character → series → appearance → composition → environment → style → quality) and fill missing essentials. Returns |
| Discover checkpoints in |
| Generate an image. Pass |
Typical agent flow: optimize_prompt → validate_prompt → generate_image.
generate_image example{
"prompt": "1girl, solo, hatsune miku, vocaloid, long hair, twintails, smile, upper body, city night, neon lights, masterpiece, best quality, very aesthetic, absurdres",
"width": 832,
"height": 1216,
"steps": 28,
"guidance_scale": 5.0,
"seed": 42
}Add "image_path": "C:/path/to/source.png", "strength": 0.5 for image-to-image (output size follows the source).
Add "loras": ["my_style.safetensors"], "lora_scales": [0.8] to apply LoRAs — they are applied per call; a call without loras runs on the bare checkpoint.
Set "render_type": "gpu" to abort instead of silently falling back to a slow CPU render when CUDA isn't available.
Base model: cagliostrolab/animagine-xl-4.0, fetched from HuggingFace (cached in ~/.cache/huggingface).
Custom checkpoints: drop SDXL .safetensors files into checkpoints/ and reference them by filename.
LoRAs: drop .safetensors files into loras/. Multiple LoRAs can be combined with independent scales.
Images are written to outputs/YYYY-MM-DD/anime_HHMMSS.png with a sidecar .json containing prompt, negative prompt, seed, size, steps, guidance, checkpoint, LoRAs and render type — enough to reproduce the image.
Animagine XL 4.0 expects Danbooru-style comma-separated tags:
End with quality tags: masterpiece, best quality, very aesthetic, absurdres.
Start with subject count (1girl, 1boy, 2girls, ...).
Character tags should be followed by their series tag.
Aim for 8+ tags; order: subject → character → series → appearance → composition → environment → style → quality.
Use the default negative prompt unless you have a reason not to.
validate_prompt and optimize_prompt enforce these for you.
pip install -e .
python -c "from anime_diffusion_mcp.server import mcp; print(mcp.name)"Package layout:
src/anime_diffusion_mcp/
├── server.py # FastMCP tools
├── prompt/ # tokenizer, classifier, validator, optimizer
├── diffusion/ # ImagePipeline (Diffusers wrapper, checkpoint/LoRA handling)
└── contracts/ # Pydantic schemas and error codesMIT — see LICENSE.
Model by Cagliostro Research Lab. Built with FastMCP and Diffusers.
Explain what each tag in a prompt does.
Breaks down the prompt into individual tags with:
Category classification (quality, composition, character, etc.)
Explanation of what each tag affects
Canonically ordered version of the prompt
| Name | Required | Description | Default |
|---|---|---|---|
| prompt | Yes | The prompt to explain |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It transparently discloses the tool's behavior: breaking the prompt into tags, classifying them, explaining their effects, and producing a canonical ordering. This gives the agent a clear picture of what happens, though it doesn't explicitly state that it's a safe, read-only operation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and well-structured: a one-sentence purpose statement followed by a tight bullet list of output components. Every line adds value without redundancy or excess length.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's low complexity (one parameter, no annotations, output schema exists), the description covers all necessary aspects: the input prompt, the breakdown process, and the key output elements (classification, explanation, canonical order). The presence of an output schema relieves the description from specifying return format, so this is complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The single parameter 'prompt' is fully described in the schema with 'The prompt to explain' (100% coverage). The description adds context about how the prompt is processed (broken into tags) but does not add new format or constraint details beyond the schema, so baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb ('Explain') and identifies the resource ('each tag in a prompt'), then elaborates with concrete deliverables (category classification, explanation, canonical ordering). This clearly distinguishes it from siblings like validate_prompt and optimize_prompt.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for understanding prompt tags but does not explicitly state when to use this tool versus alternatives like validate_prompt or optimize_prompt. No exclusions or alternative recommendations are provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Generate an image with Animagine XL 4.0.
Uses the Diffusers pipeline with the lpw_stable_diffusion_xl custom pipeline. Images are saved to outputs/YYYY-MM-DD/ with accompanying metadata JSON.
Supports custom checkpoints and LoRA mixing for style control.
Recommended workflow:
list_models → see available checkpoints and LoRAs
validate_prompt → check for issues
optimize_prompt → improve structure
generate_image → create the image
| Name | Required | Description | Default |
|---|---|---|---|
| seed | No | Random seed for reproducibility (random if not set) | |
| loras | No | List of LoRA filenames to apply (in order). Examples: ["custom_lora.safetensors"] (user's local LoRAs) | |
| steps | No | Inference steps (default 28, use 4-8 with LCM LoRA) | |
| width | No | Image width (default 832, portrait) | |
| height | No | Image height (default 1216, portrait) | |
| prompt | Yes | The positive prompt (pre-validated recommended) | |
| checkpoint | No | Checkpoint filename or 'default' for HuggingFace model. Examples: "custom_checkpoint.safetensors" (user's local checkpoint) | |
| lora_scales | No | Scale/strength per LoRA (0.0-2.0, defaults to 1.0 for each). Example: [0.8, 0.5] for two LoRAs | |
| render_type | No | Optional render type specification ('gpu' or 'cpu'). If specified and doesn't match detected device, renders are aborted to prevent slow processing. | |
| guidance_scale | No | CFG scale (default 5.0, use 1.5 with LCM LoRA) | |
| negative_prompt | No | Optional; defaults to standard negative prompt |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden of behavioral disclosure. It reveals that images are saved to outputs/YYYY-MM-DD/ with metadata JSON and that it uses a custom Diffusers pipeline, which are useful side effects. It does not describe failure modes or rate limits, but these are partially addressed by the schema (e.g., render_type).
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is front-loaded with a clear purpose, followed by relevant technical details and a valuable workflow. Each sentence contributes meaning, though the middle section about the pipeline and file output could be slightly tighter without losing key information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the complexity of an 11-parameter image generation tool, the description covers the model, pipeline, file output, and recommended workflow, making it quite thorough. The presence of an output schema handles return values. Minor gaps remain around explicit error handling and resource costs, but overall it is complete enough for an agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so baseline 3 is appropriate. The description adds context about custom checkpoints and LoRA mixing, but these already map directly to schema parameters. It does not enrich parameter meaning beyond what the schema already provides.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description opens with 'Generate an image with Animagine XL 4.0', which is a specific verb and resource, clearly distinguishing this tool from siblings like generate_image_from_image. It also states the model and pipeline used, making the tool's main function unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides a recommended workflow with explicit steps (list_models → validate_prompt → optimize_prompt → generate_image), giving clear guidance on when to invoke this tool. However, it does not explicitly state when not to use it or mention alternative tools for similar tasks.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Generate an image using img2img (image-to-image) transformation.
Takes an existing image and transforms it based on the prompt while preserving structure according to the strength parameter.
Use cases:
Style transfer (apply anime/comic/realistic style to photo)
Image refinement (improve details, fix artifacts)
Pose/composition preservation (keep layout, change style)
Character consistency (transform existing character art)
| Name | Required | Description | Default |
|---|---|---|---|
| seed | No | Random seed for reproducibility (random if not set) | |
| loras | No | List of LoRA filenames to apply (in order) | |
| steps | No | Inference steps (default 28, use 4-8 with LCM LoRA) | |
| prompt | Yes | The positive prompt describing desired output | |
| strength | No | Denoising strength (0.0-1.0). Controls how much to change. - 0.0-0.3: Minor refinements, preserve most details - 0.3-0.5: Moderate changes, good for style transfer - 0.5-0.7: Significant changes, keeps composition - 0.7-1.0: Major transformation, only basic structure preserved | |
| checkpoint | No | Checkpoint filename or 'default' for HuggingFace model | |
| image_path | Yes | Absolute path to source image to transform | |
| lora_scales | No | Scale/strength per LoRA (0.0-2.0, defaults to 1.0) | |
| render_type | No | Optional render type specification ('gpu' or 'cpu'). If specified and doesn't match detected device, renders are aborted to prevent slow processing. | |
| guidance_scale | No | CFG scale (default 5.0) | |
| negative_prompt | No | Optional; defaults to standard negative prompt |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the burden of behavioral disclosure. It explains the core transformation behavior and the role of the strength parameter in preserving structure. However, it does not mention side effects, resource requirements, or how the original image is handled. It provides moderate transparency without contradictions.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and well-structured. It opens with a clear definition, follows with a mechanistic explanation, and then lists use cases in bullet form. Every sentence adds value, and the key purpose is front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description is complete enough for a complex tool with an output schema (return values not needed). It covers purpose, transformation behavior, and use cases. Missing details like explicit alternative guidance or prerequisites are minor gaps, but the provided context is substantial.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so all parameters have descriptions. The tool description adds some context about strength preserving structure, but it does not add meaningful semantics beyond the schema. It does not explain the interplay of parameters like loras with strength or steps, so it stays at the baseline.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's function: 'Generate an image using img2img (image-to-image) transformation.' It specifies the resource (an existing image) and the transformation based on prompt and strength. This distinguishes it from the sibling tool generate_image (which is presumably text-to-image).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides a clear context for use with four explicit use cases (style transfer, image refinement, pose/composition preservation, character consistency). However, it does not explicitly state when not to use this tool or name alternatives, though the distinction from generate_image is implied by 'takes an existing image.'
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List available checkpoints and LoRAs for image generation.
Returns all available models with metadata:
checkpoints: Base models (Animagine XL)
loras: Style modifiers and speed optimizations
Use this to discover what models are available before generation.
Returns: Dictionary with checkpoints, loras, default_checkpoint, and currently_loaded
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the burden of disclosing behavior. It explains the return format: 'Dictionary with checkpoints, loras, default_checkpoint, and currently_loaded' and provides examples of content types. It also implies a read-only operation, though it does not explicitly state side-effect-free behavior.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is well-structured with a clear opening, bullet list, usage tip, and return summary. It is slightly redundant between the first sentence and the 'Use this to discover' line, but overall it is efficiently composed.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple no-parameter listing tool, the description covers the purpose, usage context, and output shape. An output schema exists, so the description does not need to detail return values further; it is complete within the scope of the tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has zero parameters and the schema is fully covered (100%), so the description has no parameters to explain. The baseline for 0 params is 4, and the description adds no unnecessary parameter info.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'List available checkpoints and LoRAs for image generation', which is a specific verb+resource. It distinguishes the tool from siblings like load_checkpoint or generate_image by focusing on discovery rather than loading or generating.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It gives explicit usage context: 'Use this to discover what models are available before generation.' This tells the agent when to invoke this tool (prior to generation) and implies it is complementary to loading/generating siblings, though it does not explicitly exclude alternatives.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Pre-load a checkpoint into GPU memory.
Loading a checkpoint in advance speeds up subsequent generation calls. Use list_models() to see available checkpoints.
| Name | Required | Description | Default |
|---|---|---|---|
| checkpoint | No | Filename from checkpoints/ folder (e.g., "custom_checkpoint.safetensors"). Use 'default' or None for Animagine XL 4.0 from HuggingFace. |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description must disclose behavioral traits itself. It mentions the side effect of loading into GPU memory and the benefit of speed, but doesn't cover failure modes, memory implications, or whether it replaces an existing checkpoint. This is acceptable but not comprehensive.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two short sentences with the action stated up front. No filler or redundant information. Every sentence contributes value: the first states the purpose, the second explains the benefit and gives a practical hint.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with a single optional parameter and an output schema present, the description is nearly complete. It covers what the tool does, why to use it, and how to find valid inputs. It doesn't mention that pre-loading might be optional or that generation may auto-load, but this is a minor gap given the tool's simplicity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema provides 100% coverage of the parameter, including default value and examples. The description adds no additional parameter semantics beyond what the schema already documents. The pointer to list_models() is helpful but not required.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's action: 'Pre-load a checkpoint into GPU memory.' This is a specific verb+resource combination that distinguishes it from siblings like list_models or generate_image. No other sibling tool performs checkpoint loading, so there is no ambiguity about its purpose.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It explicitly says loading in advance speeds up generation calls, which implies it should be used before generate_image. It also directs users to list_models() to discover valid checkpoints, providing practical usage guidance. It doesn't mention when not to use it, but for a simple pre-loading tool, this context is sufficient.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Optimize a prompt for Animagine XL.
Provide either a natural language description or an existing prompt. The optimizer will:
Reorder tags by canonical category order
Move quality tags to the end
Add missing essential categories (composition, environment, quality)
| Name | Required | Description | Default |
|---|---|---|---|
| prompt | No | Existing tag-based prompt to optimize | |
| description | No | Natural language description to convert to tags |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden and explicitly discloses the optimizer's behavior (reordering tags, moving quality tags, adding missing categories). It doesn't cover edge cases like both inputs being provided, but the primary transformations are transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is efficiently structured: a one-line purpose, a one-line input instruction, and a concise bullet list of actions. Every element earns its place, and the purpose is front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With an output schema present and only two optional parameters, the description covers the essential context: input modes and processing steps. It omits potential conflict behavior (e.g., if both inputs are provided), but this is a minor gap given the tool's simplicity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, but the description adds value by clarifying that 'prompt' and 'description' are alternative inputs ('Provide either...') and by explaining how each feeds into the optimization process. This goes beyond the schema's simple field descriptions.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states a specific action ('Optimize a prompt') and the target model ('Animagine XL'), then lists concrete transformation steps. This differentiates it from siblings like validate_prompt and explain_prompt by specifying exactly what optimization entails.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives clear input options ('Provide either a natural language description or an existing prompt') and implies the use case of improving prompt structure. It doesn't explicitly mention when not to use it or alternatives, but the context is clear enough.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Unload all LoRA weights from the current pipeline.
Useful to reset to base checkpoint style without reloading the full model. This is faster than reloading the checkpoint.
Returns: Status with success, unloaded_count, and message
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It discloses the scope ('all LoRA weights'), the purpose (reset to base style), a performance characteristic (faster than reload), and the return payload (success, unloaded_count, message). It could further elaborate on edge cases (e.g., behavior when no LoRA is loaded), but otherwise provides solid behavioral disclosure.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely concise, using only three short paragraphs. The primary action is front-loaded in the first sentence, followed by use-case justification and a brief return specification. Every sentence earns its place, with no fluff or redundancy.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
This is a simple, zero-parameter tool, and the description covers all essential aspects: purpose, when to use, performance advantage, and return values. The presence of an output schema handles the return structure, so the description's summary is sufficient for an agent to select and invoke the tool correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are zero parameters, so the description need not explain any. Per the rubric, 0 params yields a baseline of 4, and the description adds no conflicting info. The schema coverage is 100% vacuously, so no additional parameter clarification is needed.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool action with a specific verb ('Unload') and resource ('all LoRA weights from the current pipeline'). It unambiguously distinguishes itself from siblings like load_checkpoint and generate_image by focusing on removing LoRA weights.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly explains when to use it ('reset to base checkpoint style') and why it's preferable to an alternative ('faster than reloading the checkpoint'). Though it doesn't name the specific sibling tool, the comparison to reloading is concrete and actionable, giving strong contextual guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Validate a prompt against Animagine XL rules.
Checks for:
Required quality tags (masterpiece, best quality, etc.)
Proper tag ordering (quality tags at end)
Minimum tag count (8+ recommended)
Character/series consistency
Resolution compatibility
| Name | Required | Description | Default |
|---|---|---|---|
| width | No | Target image width (default 832) | |
| height | No | Target image height (default 1216) | |
| prompt | Yes | The prompt to validate | |
| negative_prompt | No | Optional negative prompt to check |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description must carry the full burden. It lists the checks performed, which gives some behavioral insight, but does not disclose whether the tool is read-only, what it returns (e.g., pass/fail, issues list), or any side effects. The presence of an output schema is noted but its content is not described. More detail on output behavior would improve transparency.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and well-structured: a single opening sentence with the verb and resource, followed by a bulleted list of checks. Every line adds value, and the content is front-loaded with the core purpose immediately.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has 4 parameters and an output schema, the description covers the main validation checks but lacks guidance on when to use it relative to siblings and does not describe expected output behavior (though output schema exists). No annotations add further gaps. It is adequate but not rich enough for complete agent decision-making.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so baseline is 3. The description adds context about prompt validation rules, which complements the prompt parameter, and mentions resolution compatibility, which relates to width/height. However, it does not add detailed semantics beyond the schema for negative_prompt or width/height.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Validate a prompt against Animagine XL rules.' It lists specific checks (quality tags, ordering, tag count, consistency, resolution), making the scope precise. This distinguishes it from sibling tools like optimize_prompt (which improves) and explain_prompt (which explains).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage by listing validation checks, but it does not explicitly say when to use this tool versus alternatives. For instance, it does not mention using this before generation or that optimize_prompt is for adjustments. No explicit when/when-not guidance is provided, only implied context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0explain_promptgenerate_imagegenerate_image_from_imagelist_modelsload_checkpointoptimize_promptunload_lorasvalidate_promptScored across 8 tools
Each tool targets a distinct operation: prompt validation, optimization, explanation, model listing, checkpoint loading, LoRA unloading, text-to-image, and image-to-image. No two tools overlap in purpose, and the descriptions make the boundaries clear.
All tools use a consistent verb_noun pattern in snake_case (validate_prompt, list_models, generate_image). The only slightly longer name is generate_image_from_image, but it follows the same convention clearly.
Eight tools cover the core workflow of prompt preparation, model management, and generation. This is a well-scoped number that avoids unnecessary redundancy.
The server covers the main image generation workflow, including prompt handling and model configuration. Minor gaps exist, such as no explicit checkpoint unload tool, but list_models includes currently loaded state and unload_loras provides a reset path.
MCP server for Flux AI image generation
MCP server for Qwen Image 3 AI image generation
MCP server for NanoBanana AI image generation and editing
MCP server for Luma Dream Machine AI video generation