mcp.film

mcp.film / Image Generation / mediagen

mediagen

Communityby Community (Cripacx)freeverified Oct 16, 2026

One MCP across Gemini, OpenAI and Kie AI, with EU AI Act disclosure marking built in.

What it does

A single Node package that ships as a CLI, a Claude Code skill and an MCP server, fronting Google Gemini (Nano Banana stills up to 4K, plus video), OpenAI's gpt-image and DALL-E, and roughly thirty Kie AI models including Flux, Imagen and Grok. The MCP surface is deliberately narrow — one generate call, one model list, one config check — so an agent picks a model by name rather than learning three vendor APIs. What sets it apart for anyone shipping into Europe is the marking step: it writes IPTC/XMP DigitalSourceType into outputs and can composite the official EU AI Act label onto a frame, which no other server in this directory does.

Connect

Claude Code
claude mcp add mediagen --env GEMINI_API_KEY=your-api-key-here -- npx -y mediagen mcp
Claude Desktop (claude_desktop_config.json)
{
  "mcpServers": {
    "mediagen": {
      "command": "npx",
      "args": [
        "-y",
        "mediagen",
        "mcp"
      ],
      "env": {
        "GEMINI_API_KEY": "YOUR_KEY_HERE"
      }
    }
  }
}
Cursor (.cursor/mcp.json)
{
  "mcpServers": {
    "mediagen": {
      "command": "npx",
      "args": [
        "-y",
        "mediagen",
        "mcp"
      ],
      "env": {
        "GEMINI_API_KEY": "YOUR_KEY_HERE"
      }
    }
  }
}
Auth: api_key · env GEMINI_API_KEY — Bring your own provider keys — GEMINI_API_KEY is the minimum, OPENAI_API_KEY and KIE_API_KEY are optional; `mediagen init` walks through configuration

Sample tools

generate_medialist_modelscheck_configuration

Field notes

Unproven and quiet: the repo was created 22 Aug 2026, its commits on main all landed inside a single day, and main has not moved since 23 Aug. Re-checked 16 Oct 2026: npm mediagen is still 0.3.0 (23 Aug), 52 stars, MIT. The only sign of life is an unmerged branch and PR by the maintainer, opened 7 Oct, that adds Kie AI video models — until it ships, what is below holds. Listed for the marking feature, not for its track record — check the commit log before you build a pipeline on it. The disclosure marking is CLI-only (`mediagen mark`) and the docs state it is never automatic, so an agent generating through the MCP tools gets no provenance metadata unless something calls the CLI afterwards. Video is Gemini-only; OpenAI and Kie AI cover stills and edits. The package is free but every generation bills the provider key you supply, and Kie AI is itself an aggregator — if you only want breadth of models, the kie-ai or fal entries here are the more direct route.

Also in Image Generation

Pairs well with