# VideoGen MCP [Verified owner] [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/video-gen/videogen-mcp  
**GitHub Stars:** 4  
**npm Downloads (last month):** 453  
**Views:** 16  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/videogen-mcp

## Description
Official VideoGen MCP server for creating product demos, ecommerce ads, and marketing videos from AI assistants. Generate and edit videos, images, voiceovers, music, and avatars; upload assets, remix projects, and export finished videos. Tools include script_to_video, storyboard_to_video, generate_image, text_to_speech, and export_project. Connect remotely at https://mcp.videogen.io/mcp with OAuth or your VideoGen API key, or run npx -y @videogen/mcp locally. Setup: https://videogen.io/videogen-mcp. Developer docs: https://docs.videogen.io/libraries/mcp. Callable tool reference: https://github.com/video-gen/videogen-mcp/blob/main/TOOL_SCHEMAS.md. Complete input and output JSON schemas for 50 hosted tools: https://raw.githubusercontent.com/video-gen/videogen-mcp/main/tool-schemas.json.

## Tools
Capabilities this server exposes over MCP:

- **get_getting_started_guidance** — How to authenticate, verify with get_me, introduce VideoGen with workflow example prompts, choose workflows vs media tools, and follow VideoGen id conventions. Call when connecting, onboarding, the user asks what VideoGen can do, or how to set up the API or MCP. Equivalent to reading the `guidance://getting-started` MCP resource — provided as a tool for clients that don't read resources directly.
- **get_async_tasks_guidance** — How to handle async workflows, tool executions, and exports: statuses, hosted vs local wait caps, polling with get_* tools, and when webhooks apply outside MCP. Call before polling or when a start tool returns a still-running snapshot. Equivalent to reading the `guidance://async-tasks` MCP resource — provided as a tool for clients that don't read resources directly.
- **get_workflows_guidance** — Canonical run → remix → export flow and when to use each workflow tool. Prefer script_to_video for longer narrated / informational videos; keep storyboard_to_video to ≤3 scenes unless the user asks for more. Call before starting a full video project. Equivalent to reading the `guidance://workflows` MCP resource — provided as a tool for clients that don't read resources directly.
- **get_tools_vs_workflows_guidance** — When to use standalone media tools (automatic model routing for a single asset) versus workflows (full editable professional video). Call when choosing between generate_* tools and workflow tools. Equivalent to reading the `guidance://tools-vs-workflows` MCP resource — provided as a tool for clients that don't read resources directly.
- **script_to_video** — Preferred for narrated / informational / explainer videos from text, especially ~1 minute or longer. Turn a verbatim narration script into an editable video with AI-generated visuals and captions. Prefer this over storyboard_to_video unless the user wants a short shot-directed storyboard. For avatar narration, pass actorEntityId and optionally set avatarQuality.
- **voiceover_to_video** — Build an editable video with AI-generated visuals from an uploaded voiceover audio file.
- **slideshow_to_video** — Build an editable narrated video from an uploaded PDF or slideshow file. Upload the file first with upload_file, then pass its fileId. For avatar narration, pass actorEntityId and optionally set avatarQuality. Returns workflowRunId, projectId, and projectUrl; poll get_workflow_run for progress and downloadUrl after auto-export.
- **storyboard_to_video** — Build an editable video from an ordered storyboard (frame-by-frame shot list). To show a saved product, actor, or visual style, pass productEntityIds, actorEntityIds, or visualStyleEntityId on each scene that should use that entity's reference images (create_entity and add_entity_reference first). A prompt alone does not attach those images. Much more credit-heavy than script_to_video: use at most 3 scenes unless the user explicitly asks for more. Prefer script_to_video for ~1 minute+ narrated / informational videos. If the user has not named a workflow, ask with pros/cons before calling this.
- **prompt_to_video_clip** — Generate one short AI video clip from a prompt inside a new editable project. Returns workflowRunId, projectId, and projectUrl; poll get_workflow_run for status and downloadUrl after auto-export.
- **video_to_video** — Restyle an existing video from a text prompt inside an editable project, keeping its length and timing. Pass the video as videoFileId (from upload_file) or as a public videoUrl, which the run imports itself. Restyles up to 120 seconds per run.
- **list_workflow_runs** — Return a page of workflow runs, most recent first, with run and project ids, status, progress, errors, and export URLs when available, plus nextCursor for pagination.
- **get_workflow_run** — Return a workflow run with status, progress, error details, projectId, projectUrl, and downloadUrl when auto-export has finished. Use this to poll an existing run without starting another project.
- **cancel_workflow_run** — Request cancellation of an API-started workflow and its related generation, remix, and export tasks. This stops pending work; it does not delete the project or undo completed work. Terminal runs are unchanged. Returns workflowRunId, projectId, projectUrl, and remixActionIds as an acknowledgment; use get_workflow_run for status.
- **generate_image** — Generate an image from a text prompt, optionally conditioned on source images (image-to-image) and actor, product, or visual-style entity ids. Typically takes 15–60 seconds. Tell the user that wait up front.
- **generate_video_clip** — Generate a video clip from a text prompt, source images, source videos, spokenDialogue, or reference audio. Combining videoFileIds with startFrameFileId or imageFileIds requires quality HIGH or MAX. Reference media guides generation without guaranteeing exact reproduction. Typically takes 1–3 minutes (HIGH/MAX can be longer). Tell the user that wait up front and keep polling calmly.
- **text_to_speech** — Synthesize spoken audio from text using a catalog voice that supportsDirectToolExecution. Returns toolExecutionId on hosted clients; poll get_tool_execution for status and results containing the audio fileId, downloadUrl, and file metadata. Local calls wait for a status snapshot.
- **generate_sound_effect** — Synthesize a standalone audio effect from a description of the desired sound and timing. Returns toolExecutionId; use get_tool_execution to retrieve the generated audio file and download URL.
- **generate_music** — Generate a music track from a text prompt. Typically takes 1–5 minutes depending on track length. Tell the user that wait up front.
- **generate_motion_graphic** — Generate an animated motion graphic video from a text prompt. Best for precise text animations (typing effects, kinetic typography, lower thirds) that stock or generated footage can't express. Outputs a transparent WebM overlay by default; set transparentBackground to false for an opaque MP4. Optionally pass reference media file ids to display or animate. Typically takes 2–5 minutes because VideoGen writes animation code and then renders it; complex prompts can take longer. Tell the user that wait before starting and keep polling calmly — a healthy in-progress job is expected.
- **generate_avatar** — Generate a talking-head avatar video from an ACTOR entity and an uploaded audio file. Pass actorEntityId and optionally set avatarQuality. Typically takes a few minutes (longer for longer audio). Tell the user that wait up front. Returns toolExecutionId; get_tool_execution provides status and results with the avatar video file and download URL.
- **vectorize_image** — Create an SVG vector version of an existing raster image. Returns toolExecutionId; get_tool_execution provides status and results with the SVG fileId and downloadUrl.
- **remove_image_background** — Create a new image with the background removed. The source file is left unchanged. Returns toolExecutionId; get_tool_execution provides status and results with the output image fileId and downloadUrl.
- **remove_video_background** — Create a new video with the background removed. The source file is left unchanged. Returns toolExecutionId; get_tool_execution provides status and results with the output video fileId and downloadUrl.
- **upscale_image** — Create a higher-resolution copy of an existing image for sharper presentation. Returns toolExecutionId; get_tool_execution provides status and the output image file. The source file is unchanged.
- **upscale_video** — Create a higher-resolution copy of an existing video. Returns toolExecutionId; get_tool_execution provides status and the output video file. The source file is unchanged.
- **image_3d_effect** — Add 3D parallax motion to a still image, producing a video.
- **list_tool_executions** — List past tool executions, most recent first.
- **get_tool_execution** — Return a tool execution snapshot with status, progress, attemptIndex, error details, and results. Each result includes fileId, file metadata, downloadUrl, and thumbnailUrl when available. Use this to poll a started generation without creating another asset.
- **cancel_tool_execution** — Request cancellation of an API-started media generation or transformation. Pending work is stopped; completed files are not deleted. Any open review for a completed result is dismissed. Returns only toolExecutionId as an acknowledgment; use get_tool_execution to read status and results.
- **list_projects** — Browse projects in the authenticated team, including dashboard-created projects, most recently updated first. Returns project ids, titles, status, editor URLs, timestamps, and aspect ratios with a pagination cursor.
- **get_project** — Return a project's title, readiness status, editor projectUrl, timestamps, aspectRatio, and assistantId when available. Use the projectId for subsequent remix and export calls.
- **export_project** — Render an editable project to MP4. Returns exportId on hosted clients; poll get_project_export with the same projectId for status, progress, errors, and downloadUrl when ready. Local calls wait for an export snapshot. Renders often take a few minutes. Tell the user that wait up front.
- **get_project_export** — Return an export snapshot with status, progress, errors, and downloadUrl, exportFileId, thumbnailUrl, and file metadata when ready. Poll until status is succeeded, failed, or cancelled.
- **remix_project** — Apply curated edits to an existing project. The original project is edited in place unless saveAsNewProject is true. CONVERT_IMAGES_TO_VIDEOS generates AI video clips from every still and is expensive. Use ZOOM for cheap Ken Burns camera motion. Poll with list_project_remix_actions for status.
- **list_project_remix_actions** — Return remix actions for a project with action ids, types, status, progress, attempts, and error details, plus pagination metadata. Poll after remix_project to determine when edits are ready for export.
- **upload_file** — Upload a local file to VideoGen and wait until it is processed. Returns the file with its id (vg_file_...) and signed URLs. Use the returned fileId for voiceover_to_video, slideshow_to_video, logos, or B-roll. To upload a remote asset, download it first and pass its local path.
- **create_file_upload** — Start an upload for a large file, or when file bytes cannot be inlined. Returns { fileId, uploadUrl }. PUT the raw file bytes to uploadUrl with NO Authorization header (it is a short-lived pre-signed URL). Then call get_file with { fileId, wait: true } to wait until processing finishes, and pass the returned fileId to workflows, tools, logos, or B-roll. For small files, prefer upload_file.
- **get_file** — Fetch a file by id with freshly hydrated (non-expired) signed URLs for its thumbnail, preview, and download renditions. Set wait: true to poll until the file finishes processing — use this right after PUTting bytes to a create_file_upload URL.
- **list_files** — List files visible to the current API key.
- **list_entities** — List built-in actors, products, visual styles, and slideshow themes plus team entities (including brands, locations, and props). Built-in rows have isBuiltIn true and cannot be updated or archived. Filter with entityType when you only need one kind.
- **create_entity** — Create an ACTOR (character), PRODUCT (something the user sells or promotes), VISUAL_STYLE, SLIDESHOW_THEME, BRAND, LOCATION (place or set), or PROP (background object that is not for sale) entity. Anything the user sells is a PRODUCT, never a PROP. A BRAND accepts brandConfig with fonts, colors, and writingStyle. Returns the new entity record with entityId and references. After create, attach references with add_entity_reference (upload the file first). Slideshow themes may attach an image or a PDF / PowerPoint. Use the returned entityId as actorEntityId on generate_avatar / script_to_video / slideshow_to_video, as productEntityIds or actorEntityIds on storyboard_to_video (request and each scene that should show it), as visualStyleEntityId on each storyboard scene, or as slideshowThemeEntityId on slideshow_to_video.
- **get_entity** — Fetch one entity by id, including its reference images. Built-in catalog entities are included and have isBuiltIn true.
- **update_entity** — Overwrite supplied display name, description, or complete brandConfig fields on an existing team entity. Omitted fields stay unchanged. Returns the updated entity. Built-in entities cannot be updated.
- **archive_entity** — Soft-archive a team entity so it no longer appears in lists or pickers. Returns entityId and archived. This tool has no restore operation; retrying an archived entity can fail because it is no longer active. Built-in entities cannot be archived.
- **add_entity_reference** — Attach an uploaded file (vg_file_...) as a reference on an entity. Images work for every entity type. Slideshow themes may also attach a PDF or PowerPoint. For a BRAND, set role BRAND_LOGO for its overlay logo or BRAND_STYLE_GUIDE for its visual generation reference. Built-in entities cannot have references added. For new PRODUCT/ACTOR entities, call this right after create_entity with isDefault: true so the entity has a usable thumbnail and generation reference.
- **remove_entity_reference** — Soft-archive a reference attachment on a team entity without deleting the source file. If it was the default reference, another reference is promoted when available. Returns the entity with its remaining references. Retrying a removed attachment can fail because it is no longer active. Built-in entities cannot have references removed.
- **list_tts_voices** — Return a page of narration voices with voiceId, display name, language, accent, description, deprecation status, and supportsDirectToolExecution. Use a voice with supportsDirectToolExecution true for text_to_speech; nextCursor fetches more voices.
- **list_languages** — Return supported narration and caption languages as languageCode and English name pairs. Use these codes when selecting workflow output languages.
- **get_me** — Fetch the account and team behind the API key (`apiKeyId`, `apiKeyNickname`, `email`, `displayName`, `teamId`). Use it as a connection test.
- **get_app_deep_link** — Build a VideoGen app URL that opens a modal or navigates after the user signs in (upgrade, buy credits, enable top-ups, invite teammates, submit feedback, integrations, account settings, or a NAVIGATE destination). Prefer this when the user needs to complete something in the VideoGen UI that MCP tools cannot do inline — especially credits or plan gates. Return the url to the user so they can open it.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "videogen-mcp": {
    "command": "npx",
    "args": ["-y","@videogen/mcp"],
    "env": {
      "VIDEOGEN_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `VIDEOGEN_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What VideoGen MCP does

VideoGen MCP connects MCP-compatible AI clients to the VideoGen API. Agents can create product demos, ecommerce advertisements, marketing videos, explainers, and other media projects from scripts, storyboards, uploaded audio, PDFs, or slideshow files.

The tool surface includes video workflows, image and video generation, text-to-speech, music, sound effects, avatars, background removal, image vectorization, upscaling, and 3D motion effects. Project tools provide project metadata and shareable URLs, while run and execution tools support listing, status checks, and cancellation.

## How it works

The package wraps the official VideoGen SDK and sends authenticated requests to the live VideoGen API. It can run locally as a stdio subprocess or remotely at `https://mcp.videogen.io/mcp` over Streamable HTTP. Local execution reads `VIDEOGEN_API_KEY`; remote requests carry an API key or OAuth access token in the `Authorization: Bearer` header.

For a typical project, an agent selects a workflow, starts generation, checks the run status, makes revisions or remixes, and exports the result. `script_to_video` is intended for longer narrated or informational videos, while `storyboard_to_video` is suited to short, explicitly directed sequences and should generally be limited to three scenes. `voiceover_to_video` and `slideshow_to_video` accept uploaded source files.

Hosted long-running operations return a workflow run or tool execution identifier immediately. Agents can poll the corresponding `get_*` tool and request cancellation when needed. Local stdio execution waits for completion by default.

## Setup and configuration

Obtain a VideoGen API key from the VideoGen API settings page. For local use, configure an MCP client to launch:

```json
{
  "mcpServers": {
    "videogen": {
      "command": "npx",
      "args": ["-y", "@videogen/mcp"],
      "env": {
        "VIDEOGEN_API_KEY": "sk_videogen_live_..."
      }
    }
  }
}
```

For remote use, point an MCP client supporting Streamable HTTP to `https://mcp.videogen.io/mcp` and send credentials with each request. OAuth discovery is available on the hosted standard endpoint when configured for the deployment.

Small local assets can be uploaded through the file tools. Larger remote uploads use a short-lived upload URL, followed by a processing check using the returned file ID. The local transport can upload from a local file path.

Optional configuration includes `VIDEOGEN_BASE_URL` for changing the upstream API endpoint and remote OAuth variables for deployments that provide account linking.

## Tools and capabilities

- Create editable videos from scripts, voiceover files, PDFs, slideshows, or ordered storyboards.
- Generate short video clips, images, talking avatars, motion graphics, music, sound effects, and speech.
- Transform media with background removal, image vectorization, image and video upscaling, and 3D parallax effects.
- Upload assets and use actor, product, or visual-style entity IDs in generation requests.
- List, inspect, and cancel workflow runs and individual tool executions.
- List projects, retrieve project metadata, and obtain shareable project URLs.
- Use guidance tools for authentication, asynchronous jobs, workflow selection, and tools-versus-workflows decisions.

## Limitations and notes

Generation and rendering are asynchronous and may take from seconds to several minutes depending on the operation. Clients should tell users when a wait is expected and poll without treating a healthy in-progress job as an error.

`storyboard_to_video` can consume substantially more credits than `script_to_video`; it is intended for short, shot-directed projects. Video exports and other long-running actions also require status handling when using the hosted transport.

The remote server is stateless and does not persist the bearer credential. The credential is forwarded to the VideoGen API for the request. The hosted documentation MCP at `https://docs.videogen.io/_mcp/server` is separate: it reads API documentation but does not execute VideoGen operations.

_Full upstream README: https://allmcps.com/mcp/videogen-mcp/readme_

