Cursor video generation with the Shotmatic MCP server

Ask Cursor for a video. Approve it before it renders.

In Cursor the agent can price a video before you build a feature around it, draft variants of a product short from your landing page copy, and render the one you pick.

Create an API keySee all 23 tools
Install

Connect Cursor to Shotmatic

  1. 01

    Create an API key on your Shotmatic account page, in the Developer section. The key is shown once; scope it to what you will ask for.

  2. 02

    Add the server to mcp.json: .cursor/mcp.json in the project, or ~/.cursor/mcp.json for every project. Replace <SHOTMATIC_API_KEY> with the key.

  3. 03

    Open Cursor's MCP settings and check that shotmatic is enabled.

  4. 04

    In Cursor's MCP settings, shotmatic shows as connected with its tools listed.

mcp.jsonCursor Remote MCP
{
  "mcpServers": {
    "shotmatic": {
      "url": "https://shotmatic.com/mcp",
      "headers": {
        "Authorization": "Bearer <SHOTMATIC_API_KEY>"
      }
    }
  }
}

Add Shotmatic to Cursor with your account API key. Your videos and credits stay in your Shotmatic account.

After connecting, call get_account and read shotmatic://capabilities. Use plan_video to choose a creation path. Add write scopes only for the actions you need.

Scopes the prompts below use: sessions:read, sessions:write, videos:read, voices:read.

Prompts

Three prompts and the calls they lead to

Shotmatic's server instructions steer the agent through these calls. It may add a read, such as the capabilities catalog, or check a status more than once.

01Price a video before you build on it

Prompt

Before I add video generation to this app: what would a 45-second vertical short cost in Shotmatic with AI images at standard quality, and with a strong AI-video start? Also show my current credit balance.

Calls it leads to

  1. 1get_creation_cost_estimate(type: "video", media_type: "images_only", image_quality: "standard", duration_seconds: 45)Prices the AI-image version.
  2. 2get_creation_cost_estimate(type: "video", media_type: "strong_start", image_quality: "standard", duration_seconds: 45)Prices the version that opens with AI video.
  3. 3get_accountReturns your balance and today's spend.
You get
Two credit estimates and your balance; nothing is created.
API key scopes
videos:read

02Two variants of a product short

Prompt

Turn the hero and features sections of app/page.tsx into a 30-second product short for Instagram Reels. Make a second variant with a more direct hook, show me both previews, and render only the one I pick.

Calls it leads to

  1. 1read shotmatic://voicesPick a ready account voice; its id becomes voice_id.
  2. 2plan_video(platform: "instagram_reels", duration_seconds: 30)Plans the short and estimates its credits.
  3. 3create_sessionCreates the first draft.
  4. 4branch_sessionForks a variant with the new hook, leaving the first draft as it is.
  5. 5get_session_previewShows each draft (called once per session).
  6. 6approve_and_renderRenders only the draft you choose.
  7. 7get_render_statusFollows that render.
You get
Two reviewable drafts and one rendered short.
API key scopes
sessions:read, sessions:write, voices:read

03A short from a blog post

Prompt

Turn our latest blog post into a 30-second vertical explainer: https://example.com/blog/launch. If the link cannot be imported, I will paste the text instead.

Calls it leads to

  1. 1read shotmatic://voicesPick a ready account voice; its id becomes voice_id.
  2. 2plan_videoReads the link in the goal and returns an article-video starter payload with its estimate.
  3. 3create_sessionStarts the draft: Shotmatic reads the article and writes the script and scenes.
  4. 4get_session_previewCheck the draft against your post.
  5. 5approve_and_renderQueues the render once it reads right.
  6. 6get_render_statusFollows the render.
You get
A narrated explainer that follows your article.
API key scopes
sessions:read, sessions:write, voices:read
Real output

A video made through these tools

Every client reaches the same server, so an output looks the same whichever agent asked for it.

Paper Moon Assembly

Made on April 3, 2026 with Shotmatic's MCP session tools, on a local copy of the hosted server using the production render pipeline and an internal test account.

Script

A city of paper astronomers gathers every night to assemble a moon from glowing fragments. Crane-sized kites lift silver crescents into place. Children in raincoats deliver the final shard. When the moon clicks together, the entire city glows as if it has remembered its own sky.

Calls the run made

  1. create_sessionCreated a draft from the script with a 9:16 render.
  2. get_session_previewWaited for the preview before changing anything.
  3. update_sessionCut the caption density, keeping the vertical framing.
  4. approve_session (today: approve_and_render)Approved the revised draft into the final render.
Format
Faceless short
Aspect ratio
9:16
Length
18 s
Visuals
AI Video
Image quality
Ultra
Visual style
Hand-painted Anime
Captions
Viral
Narration
AI voice
Soundtrack
Yes

At these settings a video like this is 301 credits today. This run predates the credit ledger, so its own charge is not on record.

Next
Claude Code guideCodex guideRecipes: step-by-step agent tasks