@cloudinary/video-mcpv0.1.0

Stream live video, generate videos from images with AI, get timestamped visual transcripts of your videos, and build reusable video transformation pipelines.

Documentation All MCP Servers Package Repository

Endpoints

EndpointMethodDescription
/mcp POST Streamable HTTP transportRecommended
/sse GET SSE transportDeprecated

Server URL

https://staging-video.mcp.cloudinary.com/mcp

Configuration

Add this to your MCP client configuration. OAuth is used by default when no headers are provided.

Minimal (OAuth)
{ "mcpServers": { "@cloudinary/video-mcp": { "url": "https://staging-video.mcp.cloudinary.com/mcp" } } }
With API Key
{ "mcpServers": { "@cloudinary/video-mcp": { "url": "https://staging-video.mcp.cloudinary.com/mcp", "headers": { "cloudinary-url": "cloudinary://api_key:api_secret@cloud_name" } } } }

Authentication

MethodDescription
OAuth 2.0 Default. The OAuth flow is initiated automatically when the client connects with no auth headers.
API Key Pass cloudinary-url header, or individual cloudinary-cloud-name, cloudinary-api-key, cloudinary-api-secret headers.

Configuration Headers

Optional headers to customize server behavior. Applied at session creation time.

HeaderDescription
cloudinary-region API region: api, api-eu, or api-ap
cloudinary-tools Comma-separated list of tools to enable (e.g. search-assets,list-images)

Response Headers

API response headers are collected into _meta.headers on each tool result. The debug preset is active by default.

HeaderDescription
cloudinary-collect-headers Control which response headers are returned. Accepts:
debug — rate-limit + request ID (default)
ratelimit — rate-limit headers only
all — every response header
Or a comma-separated list of exact names, prefix:<value>, regex:<pattern>, and preset names
cloudinary-embed-headers Set to true to also include collected headers in the response body as _headers, for clients that don't surface _meta
Tool result structure
{ "_meta": { "headers": { "x-featureratelimit-limit": "5000", "x-featureratelimit-remaining": "4998", "x-featureratelimit-reset": "Thu, 13 Feb 2026 00:00:00 GMT", "x-request-id": "bfeaccc60050594832508590a358a1a4" } }, "content": [{ "type": "text", "text": "{...}" }] }

Available Tools (26)

list-canvas-applets create-canvas-applet get-canvas-applet replace-canvas-applet delete-canvas-applet list-live-streams create-live-stream get-live-stream update-live-stream delete-live-stream activate-live-stream idle-live-stream list-live-stream-outputs create-live-stream-output get-live-stream-output update-live-stream-output delete-live-stream-output generate-video-from-image get-video-generation-job list-image-to-video-prompts create-image-to-video-prompt get-image-to-video-prompt update-image-to-video-prompt delete-image-to-video-prompt analyze-video get-video-analysis-job
destructiveidempotentopenWorldreadOnly

List applets

Returns the account's saved Video Canvas applets as summaries, without their inputs, blocks, connections, or outputs.

ParameterTypeDescription
kind Abcstring

comma-separated applet kinds to return (derived, create); defaults to derived

destructiveidempotentopenWorldreadOnly

Create an applet

Creates an applet from its content (inputs, blocks, connections, and outputs). Cloudinary compiles the action blocks to validate them and returns the stored applet. For a derived applet, the compiled blocks are also saved as a named transformation.

ParameterTypeDescription
request required { }object

The applet content to create.

blocks required [ ]array

The nodes on the canvas.

connections [ ]array

The directed edges between block ports.

description Abcstring

An optional description of what the applet does.

inputs [ ]array

The applet's declared parameters.

kind Abcstring

Applet kind. Derived applets compile into a named transformation; create applets do not. Omit on update to keep the current kind. Switching to create leaves an existing named transformation as is; switching to derived creates or redefines it.

name required Abcstring

A display name for the applet.

outputs [ ]array

The applet's declared results.

destructiveidempotentopenWorldreadOnly

Get an applet

Returns a single applet with its full content, including its inputs, blocks, connections, and outputs.

ParameterTypeDescription
appletId required Abcstring

The server-generated id of the applet.

destructiveidempotentopenWorldreadOnly

Replace an applet's content

Replaces the full content of an existing applet and recompiles its action blocks to validate them. For a derived applet, its backing named transformation is also updated.

ParameterTypeDescription
appletId required Abcstring

The server-generated id of the applet.

VideoCanvasAppletContentPayload required { }object

The new applet content.

blocks required [ ]array

The nodes on the canvas.

connections [ ]array

The directed edges between block ports.

description Abcstring

An optional description of what the applet does.

inputs [ ]array

The applet's declared parameters.

kind Abcstring

Applet kind. Derived applets compile into a named transformation; create applets do not. Omit on update to keep the current kind. Switching to create leaves an existing named transformation as is; switching to derived creates or redefines it.

name required Abcstring

A display name for the applet.

outputs [ ]array

The applet's declared results.

destructiveidempotentopenWorldreadOnly

Delete an applet

Deletes an applet and its stored content. For a derived applet, the named transformation that backs it isn't deleted, so delivery URLs and SDK calls that reference it keep working. To stop delivering the transformation, delete the named transformation separately.

ParameterTypeDescription
appletId required Abcstring

The server-generated id of the applet.

destructiveidempotentopenWorldreadOnly

Get a list of live streams

ParameterTypeDescription
request { }object

destructiveidempotentopenWorldreadOnly

Create a new live stream

ParameterTypeDescription
request required { }object

embedded_captions [ ]arraynull

idle_timeout_sec #integernull

input required { }object

type required Abcstring

max_runtime_sec #integernull

name Abcstring

destructiveidempotentopenWorldreadOnly

Get a single live stream

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

destructiveidempotentopenWorldreadOnly

Update a live stream

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

LiveStreamUpdatePayload required { }object

embedded_captions [ ]arraynull

idle_timeout_sec #integernull

max_runtime_sec #integernull

name Abcstringnull

destructiveidempotentopenWorldreadOnly

Delete a live stream

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

destructiveidempotentopenWorldreadOnly

Manually activate a live stream

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

destructiveidempotentopenWorldreadOnly

Manually idle a live stream

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

destructiveidempotentopenWorldreadOnly

Get a list of live stream outputs

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

destructiveidempotentopenWorldreadOnly

Create a new live stream output

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

LiveStreamOutputCreatePayload required { }object

asset_folder Abcstringnull

name Abcstringnull

public_id Abcstringnull

stream_key Abcstringnull

type required Abcstring

uri Abcstringnull

vendor Abcstringnull

destructiveidempotentopenWorldreadOnly

Get a single live stream output

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

liveStreamOutputId required Abcstring

live stream output id

destructiveidempotentopenWorldreadOnly

Update a live stream output

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

liveStreamOutputId required Abcstring

live stream output id

LiveStreamOutputUpdatePayload required { }object

asset_folder Abcstringnull

name Abcstringnull

public_id Abcstringnull

stream_key Abcstringnull

uri Abcstringnull

destructiveidempotentopenWorldreadOnly

Delete a live stream output

ParameterTypeDescription
liveStreamId required Abcstring

live stream id

liveStreamOutputId required Abcstring

live stream output id

destructiveidempotentopenWorldreadOnly

Create a video generation job

Submits an image and prompt for video generation and returns a job you can poll for the result.

ParameterTypeDescription
request required { }object

A JSON object with the source image, prompt, and generation options.

aspect_ratio Abcstring

Video aspect ratio. auto resolves based on image dimensions (defaults to landscape). 16:9 = landscape, 9:16 = portrait.

duration #number

Video duration in seconds.

enhance_prompt ◉booleannull

Whether to let Cloudinary automatically enhance the prompt.

folder Abcstring

Path of the folder to save the generated video to, for example campaigns/spring. If the folder doesn't exist, it's created.

generate_audio ◉boolean

Whether to generate audio. Enabling audio consumes twice as many credits.

image_asset_id required Abcstring

Cloudinary asset ID of the input image for image-to-video generation.

last_frame_image_asset_id Abcstring

Cloudinary asset ID of the last frame image.

negative_prompt [ ]array

Elements the video should avoid, for example watermark or flickering. Items must not be blank. If omitted, a default list of common artifacts is applied. Send an empty array to apply none.

prompt required Abcstring

Video description prompt.

reference_image_asset_ids [ ]array

Cloudinary asset IDs of reference images.

resolution Abcstring

Video resolution. 1080p is only supported with a duration of 8 seconds.

destructiveidempotentopenWorldreadOnly

Poll a video generation job

Returns the current status of a generation job. When the status is completed, the response includes the generated video's asset details and delivery URL.

ParameterTypeDescription
jobId required Abcstring

The video generation job ID returned from the create endpoint.

destructiveidempotentopenWorldreadOnly

Get a list of image to video prompts

Returns all saved image-to-video prompts for your product environment.

ParameterTypeDescription
request { }object

destructiveidempotentopenWorldreadOnly

Create an image to video prompt

Saves a reusable prompt with a display name so you can apply it to future generations.

ParameterTypeDescription
request required { }object

A JSON object with the prompt text and its display name.

display_name required Abcstring

A human-readable name for the prompt.

prompt required Abcstring

The prompt text describing the motion or scene to generate.

destructiveidempotentopenWorldreadOnly

Get an image to video prompt

Returns a single saved prompt by its ID.

ParameterTypeDescription
promptId required Abcstring

The ID of the image-to-video prompt.

destructiveidempotentopenWorldreadOnly

Update an image to video prompt

Updates the display name and/or prompt text of a saved prompt.

ParameterTypeDescription
promptId required Abcstring

The ID of the image-to-video prompt.

ImageToVideoPromptUpdatePayload required { }object

A JSON object with the fields to update (display name and/or prompt text).

display_name Abcstringnull

A human-readable name for the prompt.

prompt Abcstringnull

The prompt text describing the motion or scene to generate.

destructiveidempotentopenWorldreadOnly

Delete an image to video prompt

Permanently deletes a saved prompt by its ID.

ParameterTypeDescription
promptId required Abcstring

The ID of the image-to-video prompt.

destructiveidempotentopenWorldreadOnly

Create an AI Video Analysis job

Submits a stored video asset for visual analysis and returns a job you can poll for the result.

ParameterTypeDescription
request required { }object

A JSON object with the video asset to analyze and optional analysis options.

video_asset_id required Abcstring

Cloudinary asset ID of the input video to analyze.

visual_transcription_prompt Abcstring

A natural-language instruction that guides the description style or focus, for example focus on describing colors. If omitted, a default prompt is used.

destructiveidempotentopenWorldreadOnly

Poll an AI Video Analysis job

Returns the current status of an analysis job. When the status is completed, the response includes the generated visual transcript file's asset details and delivery URL.

ParameterTypeDescription
jobId required Abcstring

The AI Video Analysis job ID returned from the create endpoint.