Skip to main content
Glama
Sign In
deenesjakoruzh
sdnext_mcp

sdnext_mcp

by Congenital
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Remote

sdnext-mcp

Model Context Protocol (MCP) server for SD.Next (Stable Diffusion Next) WebUI.

It wraps SD.Next's REST API (/sdapi/v1/* + /sdapi/v2/civitai/*) as 125 MCP tools so an LLM host (Claude Desktop, DSH, etc.) can drive every SD.Next usage scenario: text-to-image, image-to-image, control-guided generation, upscaling, detailer, LoRA, scripts, captioning/tagging/VQA, model & option management, server introspection, file/gallery, and CivitAI.

This is a thin HTTP client. It does not embed Python and does not launch the WebUI. You must run SD.Next yourself with --api --listen; the MCP server just talks to it over HTTP.

The core reason this exists: SD.Next's request parameters differ from A1111 (modular guidance, 1024 default canvas, built-in YOLO detailer, schedulers as a separate axis, UNet/DiT, etc.), so A1111 MCP code does not work against it. Every tool here is modeled on SD.Next's actual pydantic request fields.

Requirements

  • Node.js ≥ 18 (uses global fetch, FormData, AbortSignal.timeout)

  • A running SD.Next instance started with --api --listen (add --api-auth user:pass to enable auth)

Related MCP server: sd-api-mcp

Install

cd sdnext_mcp
npm install          # installs @modelcontextprotocol/sdk + zod

Run

Default is stdio transport (spawned by your MCP host):

# in your MCP host config
{
  "mcpServers": {
    "sdnext": {
      "command": "node",
      "args": ["/abs/path/to/sdnext_mcp/src/index.js"],
      "env": {
        "SDNEXT_BASE_URL": "http://127.0.0.1:7860",
        "SDNEXT_USERNAME": "user",
        "SDNEXT_PASSWORD": "pass"
      }
    }
  }
}

Or run the Streamable HTTP transport directly:

SDNEXT_MCP_TRANSPORT=http SDNEXT_MCP_PORT=8787 node src/index.js
# → http://127.0.0.1:8787/mcp

Configuration (env vars)

Var

Default

Meaning

SDNEXT_BASE_URL

http://127.0.0.1:7860

WebUI address

SDNEXT_API_PATH

/sdapi/v1

v1 prefix

SDNEXT_API_KEY

—

Bearer token (highest priority)

SDNEXT_USERNAME / SDNEXT_PASSWORD

—

HTTP Basic auth

SDNEXT_TIMEOUT_MS

300000

Request timeout (generation is slow)

SDNEXT_MCP_TRANSPORT

stdio

stdio or http

SDNEXT_MCP_HOST / SDNEXT_MCP_PORT

127.0.0.1 / 8787

HTTP transport bind

SDNEXT_MCP_DRY_RUN

false

1 = print requests, don't send

SDNEXT_SAVE_DIR

—

Default dir for save_dir disk writes

Tool families (125 total)

  • Generation — sdnext_txt2img, sdnext_img2img, sdnext_control, sdnext_process_image, sdnext_process_batch, sdnext_detail, sdnext_prompt_enhance

  • Enumerators — sdnext_list_samplers … sdnext_list_masking (29 list tools)

  • Image understanding — sdnext_caption, sdnext_caption_dispatch, sdnext_tagger, sdnext_vqa, sdnext_analyze

  • Models/options — refresh/select/load checkpoint, VAE, UNet, LoRA; sdnext_get_options/sdnext_set_options

  • Server — sdnext_status, sdnext_progress, interrupt/skip, restart/shutdown, log, history, storage, GPU

  • Files/Gallery — sdnext_upload, sdnext_file, delete-, browser/, sdnext_png_info

  • Control processing — sdnext_preprocess, sdnext_mask, sdnext_detect

  • Moderation — sdnext_nudenet, sdnext_prompt_lang, sdnext_image_guard, sdnext_prompt_banned

  • CivitAI — search, model/version, download queue, settings, bookmarks, banned, history (27 tools)

  • Meta — sdnext_list_tools, sdnext_call_api (generic passthrough), autocomplete file mgmt

Run sdnext_list_tools at runtime for the full catalog, or node scripts/dump_tools.mjs.

Typical workflow

sdnext_list_checkpoints   → pick a model
sdnext_list_samplers      → pick a sampler
sdnext_txt2img { prompt, sd_model_checkpoint, sampler_name, width:1024, height:1024, steps:20 }
sdnext_progress           → poll while it runs
sdnext_caption / sdnext_tagger → understand a result

Images are raw base64 PNG in and out (no data: prefix); "upload:<id>" refs are accepted where an image is an input. Pass save_dir on any generation tool to also write the result to disk.

Scripts

npm run check          # syntax-check the entry
node scripts/smoke.mjs # DRY_RUN stdio smoke test (no live server needed)
node scripts/dump_tools.mjs  # export the tool catalog

Testing

npm test              # run the unit test suite
npm run coverage       # run tests with coverage; fails if any src/*.js file is under 80% branch coverage

Layout

See ../framework/ for the architecture documentation (API facts in 01-sdnext-api/, design decisions in 02-mcp-design/). Source under src/: config.js, client.js, image.js, registry.js, app.js (builds the server and its transports), index.js (tiny CLI bootstrap), and tools/ (one file per domain).

Available Tools

125 tools
sdnext_analyzeAnalyze Image (VLM)B

Analyze an image with the VLM "analyze" task (SD.Next /sdapi/v1/analyze). Same request shape as sdnext_vqa.

Inputs: image (required), model, top_k, top_p, +12 more.

sdnext_autocompleteList Autocomplete FilesA

List local autocomplete/tag list files (GET /sdapi/v1/autocomplete).

Inputs: no named parameters.

sdnext_autocomplete_contentRead Autocomplete FileA

Get the content of a local autocomplete/tag-list file (GET /sdapi/v1/autocomplete/{name}).

Inputs: name (required).

sdnext_autocomplete_deleteDelete Autocomplete FileA
Destructive

Delete a local autocomplete/tag-list file (DELETE /sdapi/v1/autocomplete/{name}). DESTRUCTIVE.

Inputs: name (required).

sdnext_autocomplete_downloadDownload Autocomplete FileA

Download an autocomplete/tag-list file from a remote source (POST /sdapi/v1/autocomplete/{name}/download).

Inputs: name (required).

sdnext_autocomplete_remoteList Remote Autocomplete FilesA

List remote autocomplete files with update status (GET /sdapi/v1/autocomplete/remote).

Inputs: no named parameters.

sdnext_browser_filesList Output FilesB

List files in an output folder as "folder##F##file" lines (GET /sdapi/v1/browser/files?folder=...).

Inputs: folder (required).

sdnext_browser_foldersList Output FoldersA

List output/gallery folders with labels (GET /sdapi/v1/browser/folders).

Inputs: no named parameters.

sdnext_browser_thumbOutput ThumbnailC

Get a thumbnail for an output file (GET /sdapi/v1/browser/thumb?file=...).

Inputs: file (required), exif.

sdnext_call_apiGeneric API PassthroughA

Call any SD.Next REST endpoint directly (v1 or v2, any method). Path may be absolute ("/sdapi/v1/samplers") or relative to /sdapi/v1 ("samplers"). Use only when no dedicated tool fits; prefer dedicated sdnext_* tools.

Inputs: path (required), body, query, method.

sdnext_captionCaption Image (OpenCLIP)B

Generate a text caption/prompt from an image using OpenCLIP+BLIP (SD.Next /sdapi/v1/openclip POST). mode: best|fast|classic|caption|negative. Set analyze=true for medium/artist/movement/trending/flavor breakdown.

Inputs: image (required), mode, model, analyze, +8 more.

sdnext_caption_dispatchCaption (any backend)B

Unified caption dispatch (SD.Next /sdapi/v1/caption) — pick backend: "openclip" | "tagger" | "vlm" | "analyze". Returns a unified response (caption/tags/answer depending on backend).

Inputs: image (required), backend (required), mode, model, +26 more.

sdnext_civitai_add_bannedBan ModelC

Ban a CivitAI model name (POST /sdapi/v2/civitai/banned).

Inputs: name (required).

sdnext_civitai_add_bookmarkAdd BookmarkB

Bookmark a CivitAI model name (POST /sdapi/v2/civitai/bookmarks).

Inputs: name (required).

sdnext_civitai_bannedList Banned ModelsA

List banned CivitAI model names (GET /sdapi/v2/civitai/banned).

Inputs: no named parameters.

sdnext_civitai_bookmarksList BookmarksA

List bookmarked CivitAI models (GET /sdapi/v2/civitai/bookmarks).

Inputs: no named parameters.

sdnext_civitai_check_localCheck Local by HashesA

Check which of the given file hashes are already present locally (POST /sdapi/v2/civitai/check-local).

Inputs: hashes (required).

sdnext_civitai_clear_historyClear HistoryA

Clear CivitAI search history (DELETE via POST /sdapi/v2/civitai/history).

Inputs: no named parameters.

sdnext_civitai_creatorsCivitAI CreatorsC

Search CivitAI creators (GET /sdapi/v2/civitai/creators).

Inputs: page, limit, query.

sdnext_civitai_downloadCivitAI DownloadA

Queue a model file download to the models dir (POST /sdapi/v2/civitai/download). url is required; use model_id/version_id from search results to resolve paths.

Inputs: url (required), nsfw, token, folder, +9 more.

sdnext_civitai_download_cancelCancel DownloadA

Cancel a queued CivitAI download (POST /sdapi/v2/civitai/download/{download_id}/cancel).

Inputs: download_id (required).

sdnext_civitai_download_statusDownload StatusA

List queued/in-progress CivitAI downloads (GET /sdapi/v2/civitai/download/status).

Inputs: no named parameters.

sdnext_civitai_historySearch HistoryA

List CivitAI search history (GET /sdapi/v2/civitai/history).

Inputs: search_type.

sdnext_civitai_imagesCivitAI Model ImagesC

List images for a CivitAI model or version (GET /sdapi/v2/civitai/images).

Inputs: limit, model_id, model_version_id.

sdnext_civitai_legacyCivitAI Legacy SearchA

Legacy CivitAI search endpoint (GET /sdapi/v1/civitai). Use sdnext_civitai_search for the v2 API instead.

Inputs: tag, base, nsfw, sort, +6 more.

sdnext_civitai_meCivitAI MeB

Get the authenticated CivitAI user (GET /sdapi/v2/civitai/me). Requires a configured token.

Inputs: no named parameters.

sdnext_civitai_metadata_scanMetadata ScanB

Scan local model files and match CivitAI metadata (POST /sdapi/v2/civitai/metadata/scan).

Inputs: no named parameters.

sdnext_civitai_metadata_updateMetadata UpdateA

Update sidecar metadata for local models (POST /sdapi/v2/civitai/metadata/update).

Inputs: no named parameters.

sdnext_civitai_modelCivitAI ModelB

Get a single CivitAI model by ID (GET /sdapi/v2/civitai/model/{model_id}).

Inputs: model_id (required).

sdnext_civitai_remove_bannedUnban ModelA

Remove a banned model name (DELETE via POST /sdapi/v2/civitai/banned/{name}).

Inputs: name (required).

sdnext_civitai_remove_bookmarkRemove BookmarkB

Remove a CivitAI bookmark (DELETE via POST /sdapi/v2/civitai/bookmarks/{name}).

Inputs: name (required).

sdnext_civitai_resolve_pathResolve Model PathC

Resolve the local save path for a CivitAI model/version (GET /sdapi/v2/civitai/resolve-path).

Inputs: nsfw, creator, model_id, base_model, +4 more.

sdnext_civitai_searchCivitAI SearchA

Search CivitAI models (GET /sdapi/v2/civitai/search). Filters: query, tag, types, sort, period, base_models, nsfw, limit, cursor.

Inputs: tag, nsfw, sort, limit, +7 more.

sdnext_civitai_set_settingsSet CivitAI SettingsC

Update CivitAI integration settings (POST /sdapi/v2/civitai/settings): token, save_subfolder, save_subfolder_enabled, discard_hash_mismatch.

Inputs: token, save_subfolder, discard_hash_mismatch, save_subfolder_enabled.

sdnext_civitai_settingsCivitAI SettingsA

Get CivitAI integration settings (GET /sdapi/v2/civitai/settings).

Inputs: no named parameters.

sdnext_civitai_tagsCivitAI TagsC

Search CivitAI tags (GET /sdapi/v2/civitai/tags).

Inputs: page, limit, query.

sdnext_civitai_versionCivitAI Model VersionB

Get a single CivitAI model version by ID (GET /sdapi/v2/civitai/version/{version_id}).

Inputs: version_id (required).

sdnext_civitai_version_by_hashCivitAI Version by HashB

Resolve a CivitAI version by file hash (GET /sdapi/v2/civitai/version/by-hash/{hash}).

Inputs: hash (required).

sdnext_cmd_flagsCommand FlagsA

Get all command-line flags the server was started with (GET /sdapi/v1/cmd-flags).

Inputs: no named parameters.

sdnext_controlControl-guided GenerationA

Run the control-guided pipeline (SD.Next /sdapi/v1/control) with one or more control units (controlnet / t2i adapter / xs / lite / reference / ip). Returns generated images plus preprocessor output maps.

Inputs: prompt (required), control_units (required), xyz, face, +45 more.

sdnext_delete_fileDelete FileB
Destructive

Delete a file (DELETE /sdapi/v1/delete-file?file=...). DESTRUCTIVE.

Inputs: file (required).

sdnext_delete_imageDelete ImageA
Destructive

Delete a generated image file (DELETE /sdapi/v1/delete-image?file=...). DESTRUCTIVE.

Inputs: file (required).

sdnext_detailDetailer PassC

Run the YOLO detailer on a single image as a standalone operation (SD.Next /sdapi/v1/detail) — no base generation pass. All detailer_* fields fall back to global settings when omitted.

Inputs: image (required), seed, save_dir, detailer_iou, +27 more.

sdnext_detectDetect Faces/Objects (YOLO)A

Detect faces/objects in an image with YOLO (SD.Next /sdapi/v1/detect). Returns classes, labels, boxes, crops, scores.

Inputs: image (required), model.

sdnext_extra_network_detailExtra Network DetailB

Get detailed metadata for one extra network item (GET /sdapi/v1/extra-network-detail).

Inputs: name (required), page (required).

sdnext_extra_network_detailsExtra Network Details (batch)C

Batch-fetch full details for extra network items with pagination (GET /sdapi/v1/extra-network-details).

Inputs: hash, name, page, limit, +4 more.

sdnext_fileFetch Server FileA

Fetch a file from the server (GET /sdapi/v1/file?file=...) as raw content. Must be inside an allowed dir (outputs, temp, models...).

Inputs: file (required).

sdnext_get_checkpointGet Active CheckpointA

Get info about the currently loaded checkpoint (GET /sdapi/v1/checkpoint).

Inputs: no named parameters.

sdnext_get_optionsGet OptionsA

Get all current UI/server options (GET /sdapi/v1/options).

Inputs: no named parameters.

sdnext_gpuGPU InfoA

Get GPU device info (GET /sdapi/v1/gpu).

Inputs: no named parameters.

sdnext_gpu_smiGPU SMI MetricsA

Get detailed GPU metrics: name, utilization, temperature, memory, clocks (GET /sdapi/v1/gpu-smi).

Inputs: no named parameters.

sdnext_historyGeneration HistoryC

Get generation history entries (GET /sdapi/v1/history).

Inputs: id.

sdnext_image_guardImage Guard PolicyC

Run an image through a content-safety policy/guard model (SD.Next /sdapi/v1/image-guard).

Inputs: image (required), model, policy.

sdnext_img2imgImage to ImageB

Generate images from input images with denoising (SD.Next /sdapi/v1/img2img). Supports inpainting via mask (white = repaint), color correction, hires fix, control units, and all txt2img parameters. init_images are base64 (or "upload:" refs).

Inputs: prompt (required), xyz, face, mask, +74 more.

sdnext_interruptInterrupt GenerationA

Interrupt the current generation job (POST /sdapi/v1/interrupt).

Inputs: no named parameters.

sdnext_latent_historySelect Latent History EntryA

Select a latent history entry by name (POST /sdapi/v1/latents) for "Vary (Z)"-style latent reuse. List entries via sdnext_list_latents.

Inputs: name (required).

sdnext_list_checkpointsList CheckpointsA

List all registered checkpoint models with title, filename, type and hash (GET /sdapi/v1/sd-models). Use model_name or filename for sd_model_checkpoint.

Inputs: no named parameters.

sdnext_list_control_modesList Control ModesA

List valid control modes for Union/ProMax models (GET /sdapi/v1/control-modes).

Inputs: no named parameters.

sdnext_list_controlnetsList Control ModelsC

List available ControlNet models, optionally filtered by model type (GET /sdapi/v1/controlnets).

Inputs: model_type.

sdnext_list_detailersList Detailer ModelsA

List available YOLO detailer models (GET /sdapi/v1/detailers).

Inputs: no named parameters.

sdnext_list_embeddingsList EmbeddingsA

List textual-inversion embeddings loaded/skipped for the current model (GET /sdapi/v1/embeddings).

Inputs: no named parameters.

sdnext_list_extensionsList ExtensionsA

List installed extensions (GET /sdapi/v1/extensions).

Inputs: no named parameters.

sdnext_list_extra_networksList Extra NetworksA

List extra networks of a page type (lora, checkpoint, embedding, lyco, hypernetwork...) with optional filters (GET /sdapi/v1/extra-networks).

Inputs: hash, name, page, title, +2 more.

sdnext_list_face_restorersList Face RestorersA

List available face restoration models (GET /sdapi/v1/face-restorers).

Inputs: no named parameters.

sdnext_list_ip_adaptersList IP AdaptersA

List available IP adapter models (GET /sdapi/v1/ip-adapters).

Inputs: no named parameters.

sdnext_list_latentsList Latent HistoryA

List available latent history entry names (GET /sdapi/v1/latents).

Inputs: no named parameters.

sdnext_list_loaded_lorasList Loaded LoRAsA

List currently loaded LoRA names (GET /sdapi/v1/loaded-loras).

Inputs: no named parameters.

sdnext_list_lorasList LoRAsA

List all LoRA models (GET /sdapi/v1/loras). Use in prompt as lora:name:weight.

Inputs: no named parameters.

sdnext_list_maskingList Masking OptionsA

List segmentation mask models, color maps, params and types (GET /sdapi/v1/masking).

Inputs: no named parameters.

sdnext_list_openclipList OpenCLIP ModelsA

List available OpenCLIP captioning models (GET /sdapi/v1/openclip).

Inputs: no named parameters.

sdnext_list_preprocessorsList Control PreprocessorsA

List control preprocessors with groups and configurable params (GET /sdapi/v1/preprocessors).

Inputs: no named parameters.

sdnext_list_samplersList SamplersA

List available samplers and their options (GET /sdapi/v1/samplers). Use the returned names for sampler_name.

Inputs: no named parameters.

sdnext_list_schedulersList SchedulersA

List available schedulers with class and options (GET /sdapi/v1/schedulers).

Inputs: no named parameters.

sdnext_list_scriptsList ScriptsA

List available scripts with arguments (GET /sdapi/v1/scripts). Use name for script_name.

Inputs: no named parameters.

sdnext_list_stylesList Prompt StylesA

List prompt styles (GET /sdapi/v1/prompt-styles). Apply via styles: ["style name"].

Inputs: no named parameters.

sdnext_list_tagger_modelsList Tagger ModelsA

List available tagger models (wd-tagger, deepdanbooru, etc.) (GET /sdapi/v1/tagger/models).

Inputs: no named parameters.

sdnext_list_toolsList MCP ToolsA

List every tool exposed by this MCP server with its name, HTTP method, API path, and description. Use this to discover what is available before calling sdnext_call_api.

Inputs: no named parameters.

sdnext_list_unetsList UNet/DiTA

List available UNet/DiT models (GET /sdapi/v1/unets).

Inputs: no named parameters.

sdnext_list_upscalersList UpscalersA

List available upscalers with model name, path and scale (GET /sdapi/v1/upscalers).

Inputs: no named parameters.

sdnext_list_vaeList VAEsA

List available VAE models (GET /sdapi/v1/sd-vae).

Inputs: no named parameters.

sdnext_list_vqa_modelsList VLM ModelsA

List vision-language models with prompts and capabilities (GET /sdapi/v1/vqa/models).

Inputs: no named parameters.

sdnext_list_vqa_promptsList VLM Prompt TasksA

List task prompts available for a VLM (GET /sdapi/v1/vqa/prompts).

Inputs: model.

sdnext_list_wildcardsList WildcardsB

List wildcard files available for wildcard:name syntax (GET /sdapi/v1/wildcards).

Inputs: no named parameters.

sdnext_list_xyz_optionsList XYZ Grid OptionsB

List XYZ grid axis options, optionally filtered by prefix (GET /sdapi/v1/xyz-grid).

Inputs: option.

sdnext_loaded_modelsLoaded ModelsA

List all loaded models by category (pipeline, lora, upscaler, ...) with device and size (GET /sdapi/v1/loaded-models).

Inputs: no named parameters.

sdnext_lock_checkpointLock CheckpointA

Prevent the current checkpoint from being unloaded (POST /sdapi/v1/lock-checkpoint).

Inputs: no named parameters.

sdnext_logServer LogA

Get recent server log lines (GET /sdapi/v1/log).

Inputs: clear, lines.

sdnext_log_writeWrite to LogB

Append a message to the server log (POST /sdapi/v1/log).

Inputs: json, debug, error, message.

sdnext_maskGenerate Segmentation MaskA

Generate a segmentation mask for an image (SD.Next /sdapi/v1/mask). Auto-masks when no mask is given. Mask models/types from sdnext_list_masking.

Inputs: type (required), image (required), mask, model, +1 more.

sdnext_memoryMemory InfoA

Get CPU/GPU memory usage (GET /sdapi/v1/memory).

Inputs: no named parameters.

sdnext_modulesLoaded ModulesA

List loaded model modules (GET /sdapi/v1/modules).

Inputs: no named parameters.

sdnext_motdMessage of the DayA

Get the server MOTD text (GET /sdapi/v1/motd).

Inputs: no named parameters.

sdnext_network_thumbNetwork ThumbnailC

Get a thumbnail for an extra-network item (GET /sdapi/v1/network/thumb?file=...).

Inputs: file (required).

sdnext_nudenetNudeNet CensorB

Censor explicit regions in an image with NudeNet (SD.Next /sdapi/v1/nudenet). Methods: pixelate, blur, mosaic, overlay.

Inputs: image (required), score, blocks, censor, +2 more.

sdnext_options_infoOptions InfoA

Get metadata for every option: label, section, type, default, component (GET /sdapi/v1/options-info).

Inputs: no named parameters.

sdnext_platformPlatform InfoA

Get platform/hardware info (GET /sdapi/v1/platform).

Inputs: no named parameters.

sdnext_png_infoRead Image MetadataA

Extract generation parameters from a PNG image (POST /sdapi/v1/png-info). Returns raw info string, items dict, and parsed parameters (prompt, seed, sampler, ...).

Inputs: image (required).

sdnext_preprocessRun Control PreprocessorA

Run a control preprocessor (canny, depth, pose, lineart, ...) on an image and return the processed map (SD.Next /sdapi/v1/preprocess). Preprocessor names from sdnext_list_preprocessors.

Inputs: image (required), model (required), params.

sdnext_process_batchUpscale / Process BatchC

Process multiple images at once (SD.Next /sdapi/v1/process-batch). Same options as sdnext_process_image plus an images array.

Inputs: images (required), save_dir, upscaler_1, upscaler_2, +8 more.

sdnext_process_imageUpscale / Process ImageA

Process a single image with the Extras pipeline (upscalers, e.g. from sdnext_list_upscalers). SD.Next /sdapi/v1/process. resize_mode 0 = factor, 1 = target w/h.

Inputs: image (required), save_dir, upscaler_1, upscaler_2, +8 more.

sdnext_progressQueue ProgressA

Poll generation progress: progress 0-1, ETA, state, current image (GET /sdapi/v1/progress).

Inputs: skip_current_image.

sdnext_prompt_bannedCheck Banned WordsA

Check a prompt against a banned-words list and return matched words (SD.Next /sdapi/v1/prompt-banned).

Inputs: prompt (required), words.

sdnext_prompt_enhanceEnhance PromptC

Enhance/expand a prompt with an LLM (SD.Next /sdapi/v1/prompt-enhance). Type: text, image, or video.

Inputs: prompt (required), nsfw, seed, type, +22 more.

sdnext_prompt_langCheck Prompt LanguageA

Check whether a prompt uses an allowed language/alphabet (SD.Next /sdapi/v1/prompt-lang).

Inputs: prompt (required), lang, alphabet.

sdnext_refresh_checkpointsRefresh CheckpointsA

Rescan checkpoint model folders (POST /sdapi/v1/refresh-checkpoints).

Inputs: no named parameters.

sdnext_refresh_lorasRefresh LoRAsA

Rescan LoRA model folders (POST /sdapi/v1/refresh-loras).

Inputs: no named parameters.

sdnext_refresh_unetsRefresh UNetsB

Rescan UNet/DiT model folders (POST /sdapi/v1/refresh-unets).

Inputs: no named parameters.

sdnext_refresh_vaeRefresh VAEsB

Rescan VAE model folders (POST /sdapi/v1/refresh-vae).

Inputs: no named parameters.

sdnext_reload_checkpointReload CheckpointC

Reload the current checkpoint into VRAM (POST /sdapi/v1/reload-checkpoint).

Inputs: force.

sdnext_restartRestart ServerB
Destructive

Restart the SD.Next server (POST /sdapi/v1/restart). DESTRUCTIVE.

Inputs: no named parameters.

sdnext_script_infoScript InfoB

Get info for a single script (GET /sdapi/v1/script-info).

Inputs: name (required), img2img.

sdnext_select_checkpointSelect CheckpointA

Select which checkpoint model is active (POST /sdapi/v1/checkpoint). Give the checkpoint name or filename.

Inputs: checkpoint (required).

sdnext_session_startSession StartB

Get session start info (GET /sdapi/v1/start). Optionally pass the agent name.

Inputs: agent.

sdnext_set_optionsSet OptionsA

Set UI/server options at runtime (POST /sdapi/v1/options). Body: {"options": {name: value, ...}}. Discover valid keys with sdnext_get_options / sdnext_options_info.

Inputs: options (required).

sdnext_shutdownShutdown ServerA
Destructive

Shut down the SD.Next server (POST /sdapi/v1/shutdown). DESTRUCTIVE.

Inputs: no named parameters.

sdnext_skipSkip ImageA

Skip the current image in the queue (POST /sdapi/v1/skip).

Inputs: no named parameters.

sdnext_statusServer StatusB

Get server/queue status: current job, step/steps, queue, uptime (GET /sdapi/v1/status).

Inputs: no named parameters.

sdnext_storageStorage UsageA

Get storage folder usage stats (GET /sdapi/v1/storage).

Inputs: types, folder.

sdnext_taggerAnime TaggerA

Tag an image with anime tags (Danbooru-style) using wd-tagger / DeepDanbooru (SD.Next /sdapi/v1/tagger). Model names from sdnext_list_tagger_models.

Inputs: image (required), model, max_tags, threshold, +7 more.

sdnext_torchTorch InfoB

Get torch/GPU build info (GET /sdapi/v1/torch).

Inputs: no named parameters.

sdnext_txt2imgText to ImageA

Generate images from text prompts (SD.Next /sdapi/v1/txt2img). Supports samplers, modular guidance, hires fix, refiner, detailer, LoRA (prompt lora:name:weight), styles, scripts, control units, face modules, XYZ grid. Returns base64 images, info string, and echoed parameters.

Inputs: prompt (required), xyz, face, seed, +61 more.

sdnext_unload_checkpointUnload CheckpointA

Unload the currently loaded checkpoint to free VRAM (POST /sdapi/v1/unload-checkpoint).

Inputs: no named parameters.

sdnext_uploadUpload FileA

Upload a file to the SD.Next server (POST /sdapi/v1/upload, multipart form). Files land in the system temp dir (or relative to datadir when path is given). The server accepts "upload:" refs in image fields afterwards.

Inputs: filename (required), content_base64 (required), mime, path, +1 more.

sdnext_versionServer VersionA

Get SD.Next version info (GET /sdapi/v1/version).

Inputs: no named parameters.

sdnext_vqaVQA / VLM AnalysisB

Ask a vision-language model about an image (SD.Next /sdapi/v1/vqa). Question is a task name (e.g. "Short Caption", "Object Detection", "Use Prompt"). Models from sdnext_list_vqa_models, tasks from sdnext_list_vqa_prompts.

Inputs: image (required), model, top_k, top_p, +12 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Deploy Server
    One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

    TDQS

    C2.9/5.0

    Scored across 125 tools

    Disambiguation2/5

    Many tools are unique resource/action pairs, but the set contains ambiguous clusters: legacy vs v2 CivitAI search, extra_network_detail vs extra_network_details, caption vs caption_dispatch vs tagger vs vqa vs analyze, and status vs progress. With 125 tools, agents are likely to mis-select despite good individual descriptions.

    Naming Consistency3/5

    The sdnext_ prefix and snake_case are consistent, and list_* verbs are used widely. However conventions drift: some names are resource nouns (sdnext_civitai_model, sdnext_browser_thumb, sdnext_latent_history), while others are verb_objects (sdnext_add_bookmark), and several list/action pairs (bookmarks vs add_bookmark, banned vs add_banned) are not uniform.

    Tool Count1/5

    125 tools is an extreme count for an MCP server. This is effectively a 1:1 wrapper for the full REST API, not a curated tool set, and would overwhelm any agent trying to select the right tool.

    Completeness4/5

    Coverage is impressively broad: model discovery, generation, control, image processing, captioning, CivitAI downloads, options, server management, and file operations are all represented. Minor gaps exist (e.g. no dedicated prompt-style CRUD or detailed history retrieval), but the generic sdnext_call_api fallback and delete_file cover most dead ends.

    Resources

    Related MCP Connectors

    Related MCP Servers