Works in every MCP client you already use.
Popcraft MCP · livePopcraft MCPfor any AI agent
Generate video, images, music and sound effects straight from your prompts — in Claude, ChatGPT, Cursor, or any MCP client.
https://popcraft.ai/api/mcpOAuth sign-in · no API key · token auto-refresh
claude.ai (web) and Claude Desktop — the one-click connector experience.
- 01
Copy the endpoint
One URL is the whole integration — no API key, no SDK, no config beyond a paste.https://popcraft.ai/api/mcp
- 02
Add the connector
Open Settings → Connectors → Add custom connector, name it Popcraft, and paste the URL — or jump straight to the right screen:Open Claude → Connectors - 03
Approve and create
Click Connect — your browser opens the Popcraft sign-in, you press Allow, and every tool goes live. Then open a fresh chat with a brief already loaded:Start creating in Claude
Every client here connects the same Popcraft endpoint over MCP — the one-click connector is smoothest in Claude, and only step 2 differs per app.
One brief in.
A finished deliverable out.
Not a prompt box — a production loop. Your agent plans, previews the cost, calls the tools, and hands back the result, all in one conversation.
Turn this tumbler into a 15-second vertical UGC ad — hook first, end on a call to action.
I’ll lock the product with a reference, write a hook-led script, and preview the cost before spending anything.
⚙ popcraft_media_upload → popcraft_generate_video195 credits · approved
0:15Ad’s ready — hook, benefit demo and call to action in one take, captions baked in. Want a second hook to A/B?
From product shot to UGC ad.
Upload the product once, and your agent scripts a hook-led vertical ad, locks the product with a reference, and delivers a social-ready cut with captions baked in.
popcraft_media_uploadpopcraft_generate_videopopcraft_job_displayYour agent already knows the recipe.
Ask for a finished deliverable and your agent loads the matching Popcraft skill before it generates anything — the full production line, served straight from our server. Nothing to install, nothing to paste.
Skillugc-adSingle-clip UGC product ad
Turns a product photo into a vertical creator-style ad: intake, an optional locked creator identity, a four-panel storyboard sheet, a realism pass, a credit-cost preview, then one Seedance 2.0 clip with speech and sound rendered natively.
- product intake
- creator identity
- 4-panel board
- realism pass
- one clip + QA
“Make me a 15-second vertical UGC ad for my product on Popcraft — hook people in the first second.”
Try it in Claude
Skillthumbnail-proClick-winning thumbnails
Sixteen concept frameworks, one bundled question, then an eleven-block house prompt with an identity lock for supplied face photos — parallel variants, surgical touch-ups, and clean plates sized for your own title overlay.
- concept frameworks
- 11-block prompt
- parallel variants
- micro-edits
“I need a YouTube thumbnail for my video about why cheap mechanical keyboards beat expensive ones — make it pop, using Popcraft.”
Try it in ClaudeCreate with Popcraft, inside your agent.
Real jobs, run end to end through the connector. Hover to watch, hit Recreate and the whole brief lands in a fresh chat — ready to run against your own product.
Score this scene. Generate the picture — a cream vintage convertible running a cliff-top coast road at golden hour, sunlight strobing through roadside pines, the sea glittering below, one continuous tracking shot — then write a warm uplifting nu-disco bed to sit under it: soft plucky bass, mellow electric piano, a gentle four-on-the-floor groove, understated and premium, no vocals. Lay the bed against the picture on a scoring timeline with the real waveform, a running playhead and a level meter, and mark the three points where the music actually lifts.
A handheld phone-camera creator take on a sunlit stone terrace, framed landscape and filling the whole frame. A man lifts a heavy stoneware pitcher high above a tall glass of ice and pours until the tea splashes up over the rim onto the hot stone; the camera drops to the splash, then rises with him as he sets the pitcher down, picks the glass up, swings it toward the lens until it nearly fills the frame and taps it against the glass with a laugh, drinks, grins, and sets it back down on the stone. Hard late-afternoon sun, sharp olive-leaf shadows shifting across a limewashed wall, water beading on the glass and spray catching the backlight. One continuous unbroken take, no cuts.

A handheld phone-camera creator take, landscape framing, she and the butcher-block worktop fill the whole frame and both of her hands stay inside frame the entire time. She grins at the lens and starts talking. Her right hand grips the chilled unlabelled amber bottle low around the barrel exactly as in the reference — four fingers wrapped right around the glass and clearly visible, thumb across the front, the bottle's weight sitting down in her palm, the bottle tilting exactly with her wrist. Her left hand twists the cap off and it foams over hard, a plume of white foam surging up over her fingers and spilling onto the counter. Her eyes go wide, she laughs out loud and jerks back half a step, tilts the bottle upright in that same wrapped right hand, snatches a tea towel off the counter with her left hand only and swipes the spill fast twice, then wipes her left hand and drops the towel. The right hand never lets go: the bottle stays gripped and supported in that hand for the entire take, always resting in the palm with the fingers closed around it, never balanced, never floating, never unsupported, never held by fingertips alone. She raises the bottle to her mouth with that hand, takes a real sip, gives a small delighted nod to camera and turns the wet glass so it catches the window light. Warm ordinary window daylight, condensation and foam catching the light, handheld framing that reframes with her, brisk lively real-time pace. Use the reference image for the woman's exact face, skin, hair and clothing and for the kitchen she is standing in — keep her real unretouched skin texture, pores, under-eye shadow and natural teeth exactly as photographed, no beauty filter, no skin smoothing, no re-lighting of her face. One continuous unbroken take, no cuts, no transitions, one camera. Smooth continuous fluid motion at full frame rate — never stepped, never on twos, no strobing.

Take the finished 9:16 vertical of this seltzer ad and reframe it for a widescreen placement. Hand the vertical cut itself back as the video reference and re-render the spot at 16:9 in the same sunlit kitchen with the same can and the same three beats — she lifts it into frame, cracks the tab, sips and reacts, then turns the label to camera — landing each beat on the same frame as the vertical. Keep the handheld phone-camera framing, the warm window daylight and the brisk real-time pace, and open the frame wider so the worktop and the window carry the extra width instead of empty space.
Build a complete brand guidelines deck for a fictional small ceramics studio called Lumen & Clay, then animate it as one continuous film. Generate the studio photography with Popcraft — hands centring wet clay on a wheel in warm north-window daylight, a cream-glazed stoneware bowl in hard side light, a stack of unglazed terracotta plates, and a flat-lay of blank unbranded tote, tee, kraft box and cup. Then lay out and animate eight pages: cover, logo construction and lockup variants, the five-colour palette with hex values and weights, a Baskerville type specimen, three pattern colourways, a photography-direction grid, the mark applied to the merchandise, and a closing card. Warm terracotta, bone and fired-ink palette. Pages move by push and by arc-shaped colour wipes — no hard cuts anywhere.
Make a seven-panel illustrated story film about how coffee crossed the sea, with a caption on every panel and a closing title card. Generate each panel with Popcraft in warm ink-and-watercolour editorial illustration on visible paper grain — sepia, ochre and burnt-orange with muted slate shadows: hands cupping ripe cherries on a terraced hillside at dawn; beans raked in rows on a stone roof; a dhow heeling in a heavy dusk swell with spray over the bow; porters carrying sacks down a gangplank on a misty quayside; a candlelit Venetian coffee house; a copper roaster tipping beans out through a shaft of window light; and a single cup of black coffee on a worn table. Bring every panel to life as a continuous take, cut on the beat between panels, and set the captions and the closing title in a serif over the footage.
Very tight macro of a human eye filling the frame, ringed by shattered mirror shards that repeat the reflection at the edges. Extreme resolution detail: iris striations in warm amber and copper radiating from a deep pupil, a crisp catchlight, every eyelash resolved individually, visible skin texture and fine hairs. Warm copper key light with cool glints along the broken glass edges. Cinematic editorial beauty macro photography, photoreal, no text or lettering anywhere in frame. Then upscale the finished frame to 4096 px wide, and reveal the result as one slow push into the iris that cross-dissolves from the base frame to the upscaled one, so the before and after of that same upscale arrive on screen.
A vast eroded sandstone canyon at low golden sun: one continuous unbroken crane shot that keeps discovering new things. It opens low and close on the layered rock of the near cliff, then rises and cranes steadily to the right until a narrow rope-and-plank suspension footbridge is revealed spanning the whole gap, one tiny lone figure walking across it. The sun breaks through a notch in the rim and hard god-rays fan down through the dust, igniting the haze warm gold, a bird crossing far below, before the camera eases wide onto the full span and settles. Warm painterly cinematic grade, deep amber and shadow, thick atmospheric dust, immense scale. One continuous unbroken take, no cuts, a single deliberate perfectly smooth crane move. Epic cinematic establishing shot, warm editorial grade. 16:9, fifteen seconds.
Locked-off macro shot on a heavy black cast-iron skillet on dark stone. The brown butter around the seared steak foams and rolls hard, thyme sprigs shivering in it. A hand brings a copper spoon in from the right, tips the pan toward the camera and spoons the foaming butter over the steak again and again so it sheets down the crust; a bright flame flares up along the far edge of the pan and dies back; the hand lays two more thyme sprigs into the fat and they crackle and curl, then bastes twice more and lifts the spoon away. Steam and thin smoke rise steadily through one hard warm key from the left. Deep shadow, near-black background, warm editorial food-film cinematography, fine film grain. The camera never moves. One continuous unbroken take, no cuts.
Make a title sequence for a fictional short film about things made slowly. Generate the picture with Popcraft — a heavy cast-iron letterpress running in a dark printing workshop, the flywheel spinning, brass gears meshing, inked rollers sweeping over the bed and the platen closing on a blank bone sheet, dust streaming through one hard window key, warm brass and black iron against bone paper. One continuous unbroken take, no cuts, smooth continuous motion at full frame rate. Then set three type beats and a closing title card over it in a high-contrast serif, one at a time, each easing up out of the frame.
Shoot this and then build its sound. A locked-off shot from inside a canvas tent awning on a wet hillside at dusk: heavy rain sheets off the awning edge, a small brass camp stove burns blue with an enamel kettle on it throwing steam sideways, then a hand comes in from the right, lifts the kettle by its wire handle and pours a long steady stream into a speckled enamel mug before setting it back down. Warm lantern key against cold blue rain, deep shadow, water beading on the canvas, fine film grain. One continuous unbroken take, the camera never moves. Then write the sound pass that goes under it — rain drumming on taut canvas, the low hiss of the stove, the kettle coming to a hard boil, the pour rising in pitch as the mug fills, and the kettle set down on the crate — and lay every cue against the picture on the frame where it happens.
Twenty-one tools. Zero API keys.
Everything above, plus the plumbing that makes an agent reliable — cost previews, job polling, media management, and account control.
Create5
generate_videot2v · i2v · first/last framegenerate_imageevery frontier modelgenerate_audioTTS · SFX · musiceffects_showcurated video effectslist_voicesplayable voice samples
Enhance2
upscale_imageTopaz 2x/4x · nine modesupscale_video2x/4x with fps control
Media & references6
media_uploadsigned direct uploadmedia_import_urlany public URL → referencemedia_upload_widgetinteractive file pickermedia_confirmverify uploads landedshow_mediasyour media libraryshow_elementsreusable characters & styles
Jobs & catalog5
get_workflow_instructionsproduction playbooks, served liveget_workflow_bundle_fileper-step reference modulesjob_displaypoll → result URLs livemodels_explorecapabilities + credit costsshow_generationsrevisit + chain creations
Account3
balancecredits + plan before spendingshow_plans_and_creditsone-click top-uptransactionsevery charge & refund
Cost preview on every spend.
Your agent checks the price with get_cost and asks before it burns a credit.
The frontier lineup, ready to call.
Your agent picks per task with popcraft_models_explore — capabilities and credit costs included.
image
image
image
imageQuestions, answered.
Popcraft MCP uses OAuth 2.1 — the same one-click flow you know from other connectors. Your agent opens a browser window, you sign in to Popcraft and press Allow, and the connection stays live with automatic token refresh. There are no API keys to create, paste, or leak. You can see every connected agent (with where it was last used) and disconnect any of them in Settings — and changing your password disconnects them all instantly.
Claude (web and Desktop), ChatGPT, Claude Code, Cursor, Codex, and any client that speaks the Model Context Protocol over Streamable HTTP with OAuth. Popcraft publishes standard discovery metadata and supports dynamic client registration, so compliant clients connect from the URL alone.
Images (text-to-image and reference-driven), videos (text-to-video, image-to-video, first/last-frame), and audio — including voiceover (TTS), sound effects, and full music tracks. Your agent can also browse the model catalog, check your credit balance, poll running jobs, and revisit your generation history.
Yes — generations over MCP use the same credits as the web app. Every generation tool supports a get_cost preview that returns the exact credit price without submitting anything, so your agent can always tell you the cost up front. Free-plan model access rules apply over MCP exactly as they do on the web.
Images typically take a minute or two (some models, like Midjourney, can take several); video takes a few minutes depending on the model and duration. Generation is asynchronous: your agent submits a job, gets a job id instantly, and the result card updates until the result URLs are ready — no hanging connections.
In your Popcraft account. Everything your agent creates shows up in your web-app history too, so you can keep editing on the Canvas, upscale, or download from the browser. popcraft_show_generations lists your recent creations right inside the chat.
Yes. Point popcraft_media_import_url at any public image, video, or audio URL and it becomes a reference your agent can pass to generation — as a style reference, a start frame, or an image-to-video source. Generated results chain directly too: an image your agent just made can be the start frame of the next video.
No — every new Popcraft account starts with free credits, and the MCP connector works on the free tier. Some premium models are reserved for paid plans (your agent will get a clear message and can pick an alternative from popcraft_models_explore).
From brief to deliverable in one conversation.
Free credits on signup — connect in under a minute.




