CloneViral MCP: Generate AI Images and Videos From Your Assistant

CloneViral MCP connects a compatible AI assistant to creative production tools. Instead of copying prompts between a chat and a generation interface, you can ask the assistant to create an image, animate a shot, inspect a saved flow, or translate a video through a connected tool.

MCP stands for Model Context Protocol. In this setup, your assistant is the client and CloneViral is the server. The server exposes tool schemas; the assistant discovers those schemas and calls the relevant tools with your inputs.

Choose remote MCP or a local stdio server

CloneViral offers a hosted endpoint and an npm server for clients that launch local MCP processes.

Connection	Use it when	What you configure
Remote HTTP	Your client supports remote MCP	Endpoint and supported authentication
Local stdio	Your client launches MCP servers as processes	Node.js, npm package, and API key environment
Plugin	Your host supports the packaged integration	Plugin containing MCP configuration and a skill

The remote endpoint is https://www.cloneviral.ai/api/mcp. Start at the integration setup page and select your client for its current connection flow. Authentication options depend on the client and distribution; the local examples below use an API key.

Configure a remote HTTP client

Create a key at API keys. In a client that supports HTTP MCP with environment-backed headers, the configuration follows this shape:

{
  "mcpServers": {
    "cloneviral": {
      "type": "http",
      "url": "https://www.cloneviral.ai/api/mcp",
      "headers": {
        "x-api-key": "${CLONEVIRAL_API_KEY}"
      }
    }
  }
}

This is a configuration example, not a universal format. Some clients use a settings form or a different field for HTTP transport. Confirm that your client expands environment variables; otherwise use its credential settings to supply the header securely. Do not send the literal ${CLONEVIRAL_API_KEY} string as your key.

Configure a stdio client

For a client that starts a local server, install Node.js and use the npm package:

{
  "mcpServers": {
    "cloneviral": {
      "command": "npx",
      "args": ["-y", "@cloneviral/mcp-server"],
      "env": {
        "CLONEVIRAL_API_KEY": "cv_your_api_key_here",
        "CLONEVIRAL_API_URL": "https://www.cloneviral.ai"
      }
    }
  }
}

Replace the placeholder in your private local configuration, then restart the client. Keep that file out of source control. For Claude Desktop on macOS, the documented config location is ~/Library/Application Support/Claude/claude_desktop_config.json.

Verify the connection before production

Ask your assistant:

Discover the CloneViral tools and show my credit balance. Do not generate anything yet.

The assistant should discover tools and call cloneviral_get_credits. Tool names may have a client-specific prefix, so match their suffixes. If discovery succeeds but the credit call fails, check authentication and the account configuration before submitting a generation.

Generate an image and turn it into video

A practical first workflow has four steps:

1. Call cloneviral_generate_image with your scene and output format.
2. Keep the prediction ID and call cloneviral_check_generation_status with the image type until it finishes.
3. Pass the completed image URL to cloneviral_generate_video using the current schema's supported image-to-video fields.
4. Poll the new video prediction and return the finished output link.

A prompt you can use:

Create one vertical product image of a silver skincare bottle in soft morning light. Show the completed frame first. After I approve it, use that image to create one 5-second video with a slow push-in. Check my credits before production.

The assistant should inspect live schemas for model identifiers and supported parameters. A submitted job ID means the request was accepted; it does not mean the media is ready.

Director Studio, remixing, and translation

For a cinematic shot, use cloneviral_list_cinema_presets, then cloneviral_generate_cinema_frame, and finally cloneviral_animate_cinema_shot after the frame completes. This separates visual treatment from camera movement.

For remixing, start with cloneviral_analyze_video, build variant prompts with cloneviral_build_remix_variants, and generate only the variants you choose. An analysis or variant prompt is not a rendered video.

For dubbing, cloneviral_translate_video submits the source URL and target language. Keep its job ID and check it with cloneviral_check_translate_status.

Build a reusable Viral Flow

Viral Flow represents a production pipeline as connected nodes. Ask the assistant to list your flows and inspect one with cloneviral_get_flow. It can report execution order, runnable nodes, and estimated credit cost.

To build a graph, use cloneviral_create_flow, add nodes with cloneviral_flow_add_node, and wire them with cloneviral_flow_connect. Run ready nodes with cloneviral_flow_run_node and check them with cloneviral_flow_node_status. Collect output URLs through cloneviral_flow_list_assets.

Remote tools can import publicly accessible media URLs. When your source file lives on your computer, the CLI guide covers uploading it with cviral flow upload and downloading finished flow assets.

Frequently asked questions

Does connecting MCP make generation free?

No. Production tools may consume CloneViral credits. Check the balance before running a job and inspect workflow estimates where available.

Why does my assistant know the instructions but have no tools?

A skill provides workflow guidance. The client still needs a working MCP connection and authentication. The plugin guide covers a packaged installation for supported hosts.

Should I resubmit a slow generation?

Keep its prediction ID and check status. Resubmitting can create a separate paid job. If your chat cannot continue waiting, save the pending ID so you can check it later.

Connect CloneViral MCP.