Integration details
Description
Kive helps users browse saved products and creative assets, import products, choose visual styles, and create or edit on-brand product images and videos from ChatGPT.
- Integration type
- Plugin
- Verification status
- Not applicable
- Platform
- ChatGPT
- Primary Subcategory
- AI Image & Logo Generation
- Secondary Subcategories
- None listed
- Brand
- Kive
- Access
- Account required
- First tracked
- 2026-08-05
- Tool count
- 28
- Geography
- US
The Primary Subcategory used for this profile’s headline score.
Other Subcategories where the Integration is listed.
Get alerts for Kive
Get updates when Kive’s Discoverability Score or category rank changes.
ChatGPT Plugin Discovery Score
ChatGPT Plugin discovery is coming soon
ChatGPT can surface a Plugin when it matches a user's request.Your Plugin Discovery Score measures how often yours appears.
No spam. Unsubscribe any time.
What discovery looks like

Competing in ChatGPT AI Image & Logo Generation
View Category28 tools agents can invoke
Polls one Kive character creation job started by create_character_in_kive. While the job is still preparing, it returns status creating with a poll-again instruction; creating is a normal pollable state, not a failure. When ready, it returns the saved characterId for generate_product_image or generate_product_video modelIds. If it returns status job_failed, the job is terminal: stop polling this jobId, relay the root cause to the user, and retry via create_character_in_kive with a fixed input. Do not use this for product creation jobs; those are polled with recommend_studios.
check_character_creation_status
Checks one Kive image generation id without rendering a widget. A generation may contain multiple sampled outputs. Use this only for headless/status-only contexts; in chat or visual contexts, call show_kive_image instead so the final images render inline. Returns real READY/GENERATING/FAILED status and image URLs only when Kive has returned them. Keep polling until READY or FAILED; never invent outputs. Polling always terminates: a generation with no backend progress for over 15 minutes is reported as failed (status job_failed). Stop polling, report the error, and confirm before starting a replacement generation.
check_image_generation_status
Diagnostic only. Checks the server-side Kive GraphQL identity configured for this MCP server. Do not call this before normal creative flows unless another Kive tool reports an authentication error. Opening the returned Kive URL does not authenticate this MCP connection.
check_kive_auth_status
Checks one Kive video generation id without rendering a widget. Use this only for headless/status-only contexts; in chat or visual contexts, call show_kive_video instead so the final video renders inline. Returns real READY/GENERATING/FAILED status and video URLs only when Kive has returned them. Keep polling until READY or FAILED; never invent outputs. Polling always terminates: a generation with no backend progress for over 70 minutes is reported as failed (status job_failed). Stop polling, report the error, and confirm before starting a replacement generation.
check_video_generation_status
Confirms that the caller completed the product-image upload. Call this after the Kive uploadUrl PUT returns HTTP 200, or after the fallback browser widget reports Uploaded to Kive, then use the returned uploadedImageHandle with create_product_in_kive.
confirm_product_image_upload
Creates a durable Kive character-model creation job from 1-3 reference images of the same person or character. Use this when the user wants to save a person, model, mascot, or brand ambassador as a reusable Kive character; do not use it for products (create_product_in_kive) — and style models cannot be created over MCP: apply styles at generation time via styleImageUrls or an existing style model in modelIds. There is no URL-scraping source: provide exact images via any combination of imageUrl/imageUrls, Kive imageFilePath(s), and uploadedImageHandle(s) produced by create_product_image_upload_url after a successful curl upload or fallback widget upload (the upload flow serves both products and characters); the first image is the primary reference. The job creates one saved character model; to save different characters, call this once per character. After starting, poll check_character_creation_status with the returned jobId until the character is ready, then use its id in generate_product_image or generate_product_video modelIds.
create_character_in_kive
Use this when the user wants a new board in their Kive workspace, for example to start a moodboard or collect images by theme. Creates an empty board with the given title, optionally nested inside an existing parent board, and returns its boardId. Move saved media onto it with move_kive_board_items, or browse it with find_boards. Do not use this to find or browse existing boards — use find_boards — and do not create a new board when the user refers to one that already exists.
create_kive_board
Creates a Kive MCP upload URL for one user-provided product image. For a product with multiple reference images, call this once per image file and collect every confirmed handle. First try uploading the local image bytes with the returned curlCommand. If curl fails, call show_product_image_upload_widget with fallbackUploadWidget.
create_product_image_upload_url
Creates a durable Kive MCP product creation job from exactly one source family: an ecommerce product page URL, or 1-4 reference images of the same product. Images can mix direct public image URLs (imageUrl/imageUrls), Kive imageFilePath(s), and uploadedImageHandle(s) produced by create_product_image_upload_url after a successful curl upload or fallback widget upload; the first image is the primary product image. For multiple local files, run the upload flow once per image and pass all confirmed handles in uploadedImageHandles. The job creates one saved product and stores studio recommendations; to import different products, call this once per product.
create_product_in_kive
Creates a durable MCP image edit job and edits an existing image in Kive with a natural-language instruction using Kive deterministic image editing tools. Use this for changes to an existing generated or supplied image. Use generate_product_image only when creating a new product image from a saved product.
edit_image
find_boards. Lists, shows, or browses the user’s Kive boards, and the images and videos saved on one board, such as an inspiration board or moodboard. Without boardId or boardQuery it returns all workspace boards with cover previews; with boardId or a boardQuery name match it returns that board’s saved media with stable ids and public URLs. Pass assetType IMAGE or VIDEO to narrow the media type, and query to filter items by title, caption, or generation prompt. Use returned image URLs as image inputs to other Kive tools, for example a style reference for generation or a source image for prompt extraction. Use find_generations for generated media and find_products for saved products.
find_boards
find_generations. Finds, lists, shows, displays, searches, or browses the user’s existing Kive generated media in a workspace. Use this for requests like "show me my generated videos", "show my videos generated in Kive", "show my generated images", "list my generations", or "browse my recent outputs". Pass assetType VIDEO for generated videos, IMAGE for generated images, or ALL for both. This is the discovery tool for past generations; use show_kive_video or show_kive_image only after you have a specific returned generationId or URL.
find_generations
find_models. Finds, lists, shows, displays, searches, or browses the user’s non-product Kive AI models in a workspace, such as character models, style models, people, subjects, @mentions, or names like "Finn". Use this for "show me my models", "show 5 character models", and similar requests. Requires a workspaceId returned by list_workspaces; call list_workspaces first when no workspace id is known. Uses Kive Typesense search. MCP-ready non-product models include READY and IMAGE_PREP_COMPLETED. Use find_products for saved product-library requests.
find_models
find_products. Finds, lists, shows, displays, searches, or browses MCP-ready saved Kive products in a workspace. Use this for product-library and catalog requests like "show me my products in Kive", "list my perfumes", "what fragrances do I have?", "show my catalog", "find earrings", or when the user describes one saved product before generation. Requires a workspaceId returned by list_workspaces; call list_workspaces first when no workspace id is known. Uses Kive Typesense search. MCP-ready products include READY and IMAGE_PREP_COMPLETED and can be selected for generate_product_image.
find_products
Lists Kive product-shot studios usable by generate_product_image: the published Kive catalog plus the workspace's own custom studios. Requires a workspaceId returned by list_workspaces; call list_workspaces first when no workspace id is known. Without a query, returns popular catalog studios and the workspace's newest custom studios; with query, searches both by creative direction or studio name. Custom studios are returned in customStudios on the first page only; the cursor pages only the catalog studios list. Returns stable studio ids, names, thumbnails, descriptions, and reference hints. generate_product_image accepts only studio ids returned by find_studios or recommend_studios, so use this tool to resolve a studio name the user mentions into its id. recommend_studios recommends catalog studios only, so a user's custom studios come from find_studios.
find_studios
find_video_presets. Lists Kive video presets for product-video generation. Without a query, returns recent published presets; with query, searches preset names like "slow zoom", "handheld", or "product reveal". Returns stable video preset ids, titles, poster/video previews, and reference hints usable as videoPresetId or videoPresetIds in generate_product_video. Use this for product-video style/motion selection; use find_studios only for image generation.
find_video_presets
Starts or estimates Kive image generation. Every call needs at least one of: a saved product (productId or productQuery, never both), saved models in modelIds (products, characters, styles from find_models), or a brief describing the image for prompt-only generation. workspaceId must be a workspace id returned by list_workspaces. Do not use this when the user only asks to show, list, browse, or display saved products/models; call find_products or find_models first. If it returns status not_found for a product reference, do not retry the same or another description — follow the returned nextAction, usually creating the product first with create_product_in_kive. Only pass a studioId copied from a find_studios or recommend_studios result in this conversation; never guess a studio id or pass a studio name — resolve names with find_studios first — or omit studioId. Never write model names as plain @ text in brief; resolve them with find_models and pass ids in modelIds. Pass styleImageUrls to carry the visual style of reference images (like "Apply reference style" in the composer); style images and a style model are mutually exclusive, and at most one style model can apply per generation. Optional aspectRatio is a discrete enum; strongly prefer 4:5 unless the user or host asks otherwise. Optional mode selects the quality and credit tier (premium, max); strongly prefer premium unless the user or host asks otherwise. Set estimateOnly true to check credits before spending. After starting a generation in chat, call show_kive_image with generationId and workspaceId until the image outputs render.
generate_product_image
Starts or estimates Kive video generation from saved models using deterministic GraphQL generation. Every call needs at least one saved model: a product via productId or productQuery (never both), or saved models in modelIds (products, characters, styles from find_models). brief alone cannot start a video, and workspaceId must be a workspace id returned by list_workspaces. Do not use this when the user only asks to show, list, browse, or display saved products/models; call find_products or find_models first. If it returns status not_found for a product reference, do not retry the same or another description — follow the returned nextAction, usually creating the product first with create_product_in_kive. Pass videoPresetId for one preset or videoPresetIds from find_video_presets for up to 4 motion/style presets, or omit presets. Strongly prefer the defaults unless the user or host asks otherwise: 1 output, 9:16, 5s, premium mode, audio on. Supports 1-4 samples. Supported durations by mode: draft: 5s, 6s, 7s, 8s, 9s, 10s, 11s, 12s, 13s, 14s, 15s; premium: 4s, 5s, 6s, 7s, 8s, 9s, 10s, 11s, 12s, 13s, 14s, 15s; max: 4s, 5s, 6s, 7s, 8s, 9s, 10s, 11s, 12s, 13s, 14s, 15s. Set estimateOnly true to check credits before spending. After starting a generation in chat, call show_kive_video with generationId and workspaceId until the video renders.
generate_product_video
Turns one image into a descriptive Kive generation prompt, like "Generate prompt from image" in the Kive composer. Use this for image-to-prompt-to-generation loops: describe a reference image, then pass the returned prompt as the brief to generate_product_image. Costs no credits. Do not use this to caption images for the user; it writes prompts for generation.
generate_prompt_from_image
Checks whether the configured Kive GraphQL endpoint is reachable from the MCP server.
kive_graphql_status
Checks that the Kive MCP server is reachable.
kive_ping
Returns the Kive workspaces accessible to the MCP server’s configured GraphQL identity, including each workspace id and name. Use one returned workspace id directly in later Kive tool calls.
list_workspaces
Use this when the user wants to move saved images or videos from one Kive board onto another, for example while organising or tidying boards. Pass boardItemIds using the `id` field of each item in find_boards' boardItems (not `workspaceItemId`) and the targetBoardId to move them onto; items already on the target board are left in place and reported. This only moves saved board media — it does not copy, upload, generate, or delete anything. Call find_boards first to get boardItemIds and board ids.
move_kive_board_items
Polls a Kive MCP product creation job and renders its recommended studios as an inline visual contact sheet once they are ready. While the job is still preparing, it returns a NOT READY status with a poll-again instruction; the agent should keep calling it until the studios are ready. When ready, this returns productId for generate_product_image plus studio ids for the user selection. If it returns status job_failed, the job is terminal: stop polling this jobId and follow the result nextAction — when the result carries a productId, the product was already saved, so use it with find_studios and generate_product_image instead of recreating; otherwise relay the root cause to the user and retry via create_product_in_kive with a fixed input. This tool owns the studio preview contact sheet, is display-only, and does not generate the final product image. It recommends published catalog studios only; use find_studios to list or search the workspace's own custom studios.
recommend_studios
Polls a completed/generated Kive image and renders it inline in chat once it is ready — this is the tool that owns final generated and edited image cards. Use jobId plus workspaceId for generated images, generationId plus workspaceId for a Kive generation, or imageUrl for direct edit outputs. Do not use this for studio recommendation preview images; recommend_studios owns that contact sheet. While the image is still generating it returns status "NOT_READY" with a poll-again instruction; keep calling it about every 5 seconds until it returns the image. If it returns status job_failed, the generation failed or appears stalled: stop polling and follow the returned nextAction instead.
show_kive_image
Polls a completed/generated Kive video and renders it inline in chat once it is ready — this is the tool that owns final generated video cards. Use generationId plus workspaceId for a Kive video generation, or videoUrl for a direct Kive video URL. While the video is still generating it returns status "NOT_READY" with a poll-again instruction; keep calling it about every 5 seconds until it returns the video. If it returns status job_failed, the generation failed or appears stalled: stop polling and follow the returned nextAction instead.
show_kive_video
Fallback only. Renders the browser upload widget for a product image after the Kive MCP upload curlCommand from create_product_image_upload_url fails or is blocked. Do not call this before trying curl.
show_product_image_upload_widget
Use this when the user wants to rename a Kive board, change or clear its description, or move the board itself under another board or back to the workspace top level. Pass the boardId from find_boards plus only the fields to change. Nesting under a parent with a different link-sharing setting applies the parent's setting to the board and its sub-boards — the tool asks for confirmSharingChange: true before doing that. This does not move images or videos between boards — use move_kive_board_items for that — and it cannot delete boards.
update_kive_board
How do I improve a ChatGPT Plugin's discoverability?
The levers are the listing surface agents actually read: names, descriptions, keywords, tool metadata, and registry health. Which lever matters depends on where discovery breaks, which is what continuous measurement shows.
What are Kive alternatives on ChatGPT?
As of 2026-09-19, Kive competes with Alpix, Apixel, Bloom, HTML/CSS to Image API, Inkroost, LogoGenic Image Generator, MangaBoom, OEG AI Studio, Vivin in ChatGPT AI Image & Logo Generation, ranked by public Discoverability Score.
Where is this profile measured?
This profile uses the geography attached to the latest public registry snapshot: US. Locale tags are intentionally omitted.