+1 - 866 706 7890
Connect PixelBin to any MCP-compatible AI client and unlock PixelBin’s full predictions catalogue — image generation, video generation, background removal, watermark cleanup, upscaling, restoration, colourisation, and more — right from your conversation. Outputs come back as CDN URLs, with one call to persist them permanently to your PixelBin storage.
Pick your client. Every config below is copy-paste — the MCP server handles auth.
PixelBin and paste the URL below. Click Add → Connect, sign in with your PixelBin account.
Connect the PixelBin MCP at https://mcp.pixelbin.io/mcp — it supports OAuth device flow, discovery at https://mcp.pixelbin.io/.well-known/oauth-authorization-server. Run the device flow and send me the link like: Authorize here: https://console.pixelbin.io/device?user_code=XXXX-XXXX. Then poll the token endpoint until I authorize and confirm once it succeeds.
Connect the PixelBin MCP at https://mcp.pixelbin.io/mcp — it supports OAuth device flow, discovery at https://mcp.pixelbin.io/.well-known/oauth-authorization-server. Run the device flow and send me the link like: Authorize here: https://console.pixelbin.io/device?user_code=XXXX-XXXX. Then poll the token endpoint until I authorize and confirm once it succeeds.
A full PixelBin media stack — generation, cleanup, enhancement, storage — behind one MCP connection. Your AI client picks the right plugin and operation for each prompt automatically.
Image generation with nanoBanana, nanoBanana 2, and nanoBanana Pro. Video generation with Sora 2, Veo 3, Veo 3 Fast, Kling 3, Hailuo 2.3, Seedance, LTX-2, and Wan 2.5. Your AI client lists the catalogue, picks a plugin based on the prompt, and runs it as a prediction.
Every plugin advertises its credit cost via estimate-prediction-cost. Well-behaved AI clients call it first, surface the cost in chat, and only then run create-prediction. Hard credit limits live server-side — jobs that would exceed your balance are rejected before they consume the bearer.
Background removal with foreground-type selector (general, e-commerce, car, human, object). Watermark cleanup on images and videos you own or have rights to edit. Image and video upscaling at 2x, 4x, or 8x. Photo restoration, colourisation, expansion, and relighting.
Need to feed a local file into a plugin? Call request-upload-url for a signed URL, push the file, then pass the returned CDN URL to create-prediction. To keep an output forever, call save-prediction-to-storage — it lands in your PixelBin DAM with a permanent cdn.pixelbin.io URL.
A small, sharp surface. Together they cover discovery, cost estimation, execution, retrieval, upload, and persistence.
create-prediction. Use it to keep the human in the loop on spend.
request-upload-url first if the source is local. Returns a prediction id you can poll.
save-prediction-to-storage to keep them forever.
create-prediction.
cdn.pixelbin.io URLs. Choose folder path and access level (public-read or private).Five workflows where one MCP connection replaces a stack of point tools.
Generate, clean, upscale, and persist hero shots for Amazon, Shopify, and your storefront from a single AI conversation.
Remove watermarks and unwanted text from images you own or have rights to edit, then save the cleaned outputs to your DAM — all via your AI client.
Upscale catalogue and marketing images by 2x, 4x, or 8x for print, billboards, and large-format placements.
Pick a still as a keyframe, brief a clip with Veo 3, Sora 2, or Kling 3, and persist the result — all from your IDE or chat client.
Restore detail in old photographs and colourise black-and-white archives, all reachable from a chat prompt.
Six steps inside Claude, Codex, or any MCP-compatible client. Show More for steps four through six.
https://mcp.pixelbin.io/mcp into your client’s connector settings. Authenticate with your PixelBin account.
request-upload-url. Either way the image lands as a public CDN URL ready to feed into a prediction.
marketplace/ folder.” Your AI client lists predictions, picks plugins, and chains them.
estimate-prediction-cost and surfaces the credits required in chat. You confirm; the run proceeds. Out of credits? The server rejects the call before it bills.
save-prediction-to-storage. Outputs land in your DAM with permanent cdn.pixelbin.io URLs and the access level you chose.
Different briefs, different deliverables. Same MCP connection.
Add the MCP server to your AI client of choice and your assistant can generate, transform, store, and deliver media without you stitching SDKs. One config line and your conversational workflow gains a media pipeline.
Run bulk catalogue refreshes and marketplace exports straight from chat. Your AI client orchestrates list → estimate → create → save loops; the DAM keeps every output organised under your folder structure.
Concept hero shots, generate ad variants in image and video, produce social-ready cuts, and ship a campaign without leaving the AI conversation. The MCP turns your client into a creative ops layer.
One MCP, multiple clients. Use folders and tags to keep brand sets separate, signed URLs to share private drafts, and chat-driven runs to deliver a whole catalogue or campaign in one session.
For the dev or AI engineer wiring this in.
POST https://mcp.pixelbin.io/mcp. Two OAuth 2.0 flows: authorization-code with PKCE (RFC 7636) for clients that can host a callback (Claude Desktop, Claude Code), and Device Authorization Grant (RFC 8628) for chat-only or bearer-only clients (paste-prompt agents, Codex CLI, Cursor, custom MCP-SDK consumers). Dynamic client registration (RFC 7591) and discovery via /.well-known/oauth-authorization-server (RFC 8414) + /.well-known/oauth-protected-resource (RFC 9728). No API keys to copy.
tools/list + tools/call spec. Six tools: list-predictions, estimate-prediction-cost, create-prediction, get-prediction, request-upload-url, save-prediction-to-storage. Every plugin in the PixelBin predictions catalogue is reachable through create-prediction.
get-prediction; the final URL is hosted by PixelBin for ~30 days. Persist forever with save-prediction-to-storage, which returns the permanent cdn.pixelbin.io URL and the credits consumed.
estimate-prediction-cost so your AI client can surface it in chat before running. Hard credit limits live server-side — jobs that would exceed your balance fail before they bill.save-prediction-to-storage moves them into your DAM with a permanent cdn.pixelbin.io URL and the access level you choose.create-prediction.create-prediction.list-predictions, estimate-prediction-cost, create-prediction, get-prediction, request-upload-url, save-prediction-to-storage.private access flag in save-prediction-to-storage for any work you don’t want indexed before launch.request-upload-url. Raw bytes / base64 / data URLs are not accepted by create-prediction.For anything else, read the PixelBin docs.
console.pixelbin.io/device) for chat-only agents. Either way the AI client never stores a long-lived secret you minted by hand. Your existing PixelBin plan and credits work seamlessly.list-predictions to enumerate the live set with input schemas.estimate-prediction-cost first and surfaces the credit cost in chat — all major MCP clients (Claude, Cursor, Codex) follow the tool description and do this. Hard layer: the PixelBin platform rejects any prediction that would exceed your credit balance with a 402 before it bills, regardless of what the AI client did. Watch your spend in the PixelBin console; cap it via plan limits.save-prediction-to-storage — it lands in your PixelBin DAM as a cdn.pixelbin.io/v2/<your-cloud>/... URL backed by HTTP/3 edge caching, with a folder path and access level (public-read or private) of your choice.request-upload-url with the filename; the server returns a signed PUT URL plus the public CDN URL the file will land at. Upload, then pass that CDN URL to create-prediction. Supported formats: image (jpg / jpeg / png / webp / avif / gif / bmp / tiff / tif / heic / heif), video (mp4 / webm / mov / avi / mkv / mpeg / mpg), and pdf.