Model Context Protocol (MCP) Setup

Connect your favorite AI assistants to Wevi to create, customize, and render professional videos directly via natural language prompts.

1. Prerequisites & Package Info

1. An active account on app.wevi.ai.
2. A valid API Key from Profile → Developer API Keys & MCP (wevi_live_... or wevi_test_...).
3. Node.js 18+ installed on your local machine.

2. Client Configurations

A. Claude Desktop

1. Open your Claude Desktop configuration file in your text editor:

   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json

2. Add the wevi server entry:

{
  "mcpServers": {
    "wevi": {
      "command": "npx",
      "args": ["-y", "@wevi/mcp@latest"],
      "env": {
        "WEVI_API_KEY": "wevi_live_YOUR_KEY_HERE"
      }
    }
  }
}

3. Save the file and restart Claude Desktop. The hammer / tool icon in Claude will show the 10 registered Wevi tools.

B. Claude Code CLI

Add the server with a single terminal command:

claude mcp add wevi -e WEVI_API_KEY="wevi_live_YOUR_KEY_HERE" -- npx -y @wevi/mcp@latest

(This automatically saves the server and connects your API key in one step).

C. Cursor & Windsurf IDEs

Create or update .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "wevi": {
      "command": "npx",
      "args": ["-y", "@wevi/mcp@latest"],
      "env": {
        "WEVI_API_KEY": "wevi_live_YOUR_KEY_HERE"
      }
    }
  }
}

3. Prompts to try

Paste any of these into Claude Desktop, Cursor, or Claude Code once the server is connected.

• Launch video — "Using Wevi, build a 3-scene launch video for my product: a hook, one key feature, and a call to action. Put all scenes in one project, render it, and give me the MP4 link."
• Screenshot to video — "Using Wevi, capture a screenshot of https://example.com, then create a one-scene showcase video that uses it as the image layer and render it."
• Check a render — "Using Wevi, check the render status of my latest project and give me the video URL when it is ready."
• Credits — "Using Wevi, how many credits do I have left and what does a 3-scene 1080p video cost?"

4. Example Natural Language Workflows

🎥 Generating a SaaS Promo Video

User Prompt:
"Search Wevi for published templates. Build a 3-scene video for a productivity app called 'FocusFlow' (hook, feature, CTA) and render it."

What Claude executes behind the scenes:

1. Calls wevi_list_templates — only published templates come back.
2. Calls wevi_get_template_schema({ templateId }) for each chosen template to read layer keys (TXTTYPINGVAR, CLRTEXTTYPINGVAR, …).
3. Calls wevi_create_project({ title, scenes: [ { templateId, parameters, voiceoverText }, … ] }) — one project holding all scenes. The aspect ratio is derived from the templates.
4. Calls wevi_trigger_render({ projectId }) — renders every scene and exports one concatenated MP4 with music and voiceover.
5. Polls wevi_get_render_status({ projectId }) until phase is completed, then returns videoUrl.

🧪 Trying it safely

Use a wevi_test_ key while experimenting: every render is watermarked, capped at 720p, and uses no credits. Test keys are limited to 20 scene renders and 5 exports per day, and 4 scenes per project.

🌐 Website Screenshot to Video

User Prompt:
"Take a screenshot of stripe.com and use it as the background image in our product showcase video template on Wevi."

What Claude executes:

1. Calls wevi_capture_web_ui({ url: 'https://stripe.com' }) to capture a high-resolution image.
2. Passes the returned asset URL as the value of the template's image layer in wevi_create_project.
3. Renders with wevi_trigger_render and returns the MP4 download URL.