Levea MCP Server

What levea-mcp-server MCP server does

levea-mcp-server MCP server exposes Levea’s prompt-driven video editing workflow to local AI clients. The wrapper sends creative instructions and media-related requests to the hosted Levea production engine, which builds and updates structured video projects rather than returning only a flattened output.

Supported workflows include frame-accurate cutting, silence removal, kinetic captions, active-speaker reframing, chroma key processing, motion graphics, Google Cloud text-to-speech voiceovers, and cloud-rendered MP4 export. The current project description also identifies viral clip creation and caption generation as supported uses.

Projects retain a scene and timeline representation containing tracks, layers, assets, text, masks, audio, effects, and transitions. That representation can be returned as an editable scene, used to queue media work, or exported for delivery.

How it works

The MCP package is a client-side stdio wrapper. An AI editor starts it with npx, and the wrapper authenticates requests against the Levea API. The hosted service performs the resource-intensive work, including transcription, rendering, face tracking, motion graphics, and compositing, so local clients do not need a CUDA or GPU installation for those operations.

Levea separates planning from execution. A planning layer interprets prompts, transcripts, visual references, and timeline boundaries, then produces typed editing operations. A deterministic production harness applies those operations to the project’s scene graph. Verification and bounded repair are part of the described pipeline, while immutable scene payloads provide linear undo and redo history across revisions.

The result may be an updated editable scene, queued media work, an MP4, or an asset bundle for multi-platform delivery. Export is optional rather than the only output path.

Setup and configuration

To run levea-mcp-server MCP server, create a Levea API key through livecore.ai. Register the package as a command-based MCP server in the client’s MCP configuration:

{
  "mcpServers": {
    "levea": {
      "command": "npx",
      "args": ["-y", "levea-mcp-server"],
      "env": {
        "LEVEA_API_URL": "https://api.livecore.ai",
        "LEVEA_API_KEY": "your-key-from-livecore.ai"
      }
    }
  }
}

LEVEA_API_URL should contain the bare host https://api.livecore.ai; the wrapper appends the required endpoints. The README documents setup for Cursor, Windsurf, Cline, Claude Desktop, and Claude Code. Claude Code can add the server with its MCP command and environment-variable options.

Tools and capabilities

The levea-mcp-server MCP server supports agent-directed operations for:

• Creating short or viral clips from source footage
• Applying frame-accurate cuts and timeline edits
• Generating captions with kinetic typography styles
• Reframing footage around an active speaker
• Applying chroma key effects
• Generating voiceovers through Google Cloud TTS
• Producing motion graphics such as charts, cards, waveforms, and diagrams
• Returning editable scene data or exporting cloud-rendered MP4 files

The production architecture also describes silence cuts, ripple and slip/slide edits, split-screen composition, audio ducking, loudness mastering, and multi-platform asset delivery.

Limitations and notes

The open-source MCP wrapper is not the video renderer itself. It connects to Levea’s hosted production engine, so a Levea API key and network access to the configured API are required. Heavy processing is performed remotely, and the README describes free starter credits as available through livecore.ai.

The material does not document a local rendering mode, offline operation, or a complete individual tool schema. It also does not state that every listed rendering or media backend is directly exposed as a separate MCP tool. Use the server when the hosted Levea workflow and its structured scene model fit the project requirements.