SpecialX inside your AI assistant.
Plug SpecialX into Claude, Cursor, ChatGPT, Windsurf, or any MCP-compatible client. Generate images and video, manage workspaces, and check task status straight from your assistant.
Connect to Claude
Add a custom connector
Name it SpecialX and paste the URL:
https://specialx.ai/api/mcpConnect and sign in
Click Add → Connect, sign in with your SpecialX account — you're all set, now just ask Claude to generate an image.
What your assistant can do
Once connected, your client gets these 22 tools out of the box.
Generate
list_modelsLists available image and video models with their input schemas and credit costs.
generate_imageSubmits an image generation job to a chosen model and workspace. Returns a task_id.
generate_videoSubmits a video generation job (Veo, Kling, Sora, Grok, and more). Returns a task_id.
get_taskPolls the status of a generation task and returns the final media URLs when complete.
Media library
get_upload_urlReturns a presigned URL to upload your own image/video, plus the public URL to use it as a reference input.
list_mediaLists your uploaded media (or a workspace's generated outputs) to reference as inputs to generations.
Workspaces & workflows
list_workspacesReturns the SpecialX workspaces the authenticated user belongs to.
create_workspaceCreates a new workspace (project container) and returns its id and URL.
rename_workspaceRenames a workspace you own.
list_workflowsLists the user's workflow canvases — personal plus any workspace they belong to.
create_workflowBuilds a workflow canvas from a high-level blocks spec (notes + reference assets + generators), ready to Generate.
update_workflowRenames a workflow and/or appends new blocks below its current content.
get_workflowReads a single workflow canvas with its nodes (text, asset, mediaGen, llm, group) and edges.
get_group_mediaResolves a workflow group node into its media items + context notes, with Meta connection status.
Billing
billing_statusReports your plan, subscription status, and credit balance — and whether generation is enabled.
get_checkout_linkMints a Stripe checkout link for a plan or credit pack so you can pay without leaving your assistant.
Meta Ads
meta_connection_statusChecks your Meta connection and can list available ad accounts, pages, and pixels.
deploy_meta_ads_planCreates or extends a Meta campaign, ad sets, and ads from a plan (supports carousel + dry-run).
list_meta_campaignsLists campaigns under your connected ad account, newest first.
list_meta_adsetsLists the ad sets within a given Meta campaign.
manage_meta_statusSets ACTIVE / PAUSED / ARCHIVED / DELETED on a campaign, ad set, or ad.
update_meta_budgetUpdates the daily or lifetime budget (in cents) on a campaign or ad set.
Available models
Pass any model id below to generate_image or generate_video. Credit costs start from the listed base and rise with resolution, quality, or duration.
Image models
12 models| Model | model id | From credits |
|---|---|---|
GPT Image 2.5 FlareNEW OpenAI's fast image model for everyday generation and editing | openai/gpt-image-2.5-flare | 8 cr |
GPT Image 2.5 SunburstNEW OpenAI's most capable image model for precise editing and detailed imagery | openai/gpt-image-2.5-sunburst | 8 cr |
GPT Image 2NEW OpenAI's State-of-the-Art Image Generation | openai/gpt-image-2 | 2 cr |
Grok ImagineNEW xAI Grok image generation | xai/grok-imagine-image | 1 cr |
Nano Banana Pro Google's Flagship Generation Model | google/nano-banana-pro | 1 cr |
GPT Image 1.5NEW OpenAI's Latest Image Generation | openai/gpt-image-1.5 | 2 cr |
P-ImageFAST Fast & High Quality Generation | prunaai/p-image | 1 cr |
Z-Image Turbo Img2ImgFAST Fast img2img — re-imagines a sketch from your prompt | prunaai/z-image-turbo-img2img | 1 cr |
Seedream 4.5NEW ByteDance Seedream 4.5 | bytedance/seedream-4.5 | 2 cr |
Seedream 5 LiteNEW ByteDance Seedream 5.0 Lite with built-in reasoning | bytedance/seedream-5-lite | 2 cr |
Topaz Image UpscaleNEW Professional-grade image upscaling from Topaz Labs | topazlabs/image-upscale | 2 cr |
FLUX.2 ProPRO Black Forest Labs FLUX.2 Pro | black-forest-labs/flux-2-pro | 3 cr |
Video models
13 models| Model | model id | From credits |
|---|---|---|
Seedance 2.5NEW Up to 30 seconds with native audio and multimodal references · Replicate | bytedance/seedance-2.5 | 36 cr |
MiniMax H3 MaxNEW Text or first/last-frame video with native audio · fal.ai | fal-ai/minimax/h3-max | 15 cr |
MiniMax H3 Max TurboLOW COST Lower-cost fast video with native audio · fal.ai | fal-ai/minimax/h3-max-turbo | 8 cr |
Kling 3.0 · HiggsfieldNEW Kling Standard/Pro via Higgsfield; requires the platform API connection | higgsfield/kling-v3 | 27 cr |
Sync-3 AvatarNEW Fal talking portrait with exact voice-track lip-sync (up to 30 seconds) | fal-ai/sync-lipsync/v3/image-to-video | 0 cr |
Seedance 2.0NEW ByteDance multimodal video with native audio | bytedance/seedance-2.0 | 21 cr |
Grok Imagine Video 1.5NEW xAI Grok video generation (1.5 with reference image) | xai/grok-imagine-video | 18 cr |
Gemini Omni FlashNEW Google Gemini Omni Flash video generation & editing | google/gemini-omni-flash | 31 cr |
Veo 3.1NEW Google's Video Generation Model v3.1 | google/veo-3.1 | 100 cr |
Veo 3.1 LiteNEW Cost-efficient Veo 3.1 with native audio | google/veo-3.1-lite | 13 cr |
Kling 3.0 OmniNEW Kling V3 Omni with sound & image/video reference | kwaivgi/kling-v3-omni | 37 cr |
Kling 3.0NEW Kling V3 with multi-shot, elements & sound | kwaivgi/kling-v3 | 53 cr |
Kling 3.0 Motion ControlNEW Motion transfer from reference video onto a character image | kwaivgi/kling-v3-motion-control | 21 cr |
How authentication works
- 1.Your client opens a browser window to
specialx.ai/oauth/authorize. - 2.You sign in (or sign up) and approve access. Tokens are stored locally by your client.
- 3.All requests are scoped to your account. Generations consume credits from your balance — refunded automatically on failure.