Ronana CLI and MCP

Requires a Ronana paid plan.

WARNING: Generations spend your account credits.

Install

Requires Node.js 20 or later.

npm install -g @ronana/mcp

This installs two commands: ronana, the command line, and ronana-mcp, a local MCP server for AI apps.

Related MCP server: Templated MCP Server

Sign in

ronana auth login

Ronana opens in your browser. Sign in and approve the access. You do not need an API key. The sign-in is stored in your system keychain and stays until you run ronana auth logout, which also revokes it.

Command line

ronana auth login | status | logout
ronana account
ronana model list [filter] [--kind image|video]
ronana model get google/nano-banana-2-lite/edit
ronana upload ./photo.png                  # prints a ref_ id
ronana generate cost   google/nano-banana-2-lite/edit --prompt "make it dusk" --image ./photo.png
ronana generate create google/nano-banana-2-lite/edit --prompt "make it dusk" --image ./photo.png --out ./renders
ronana generate get <id> | wait <id> [--out dir]
ronana generate list
ronana upscale <id | file | url> --resolution 4k --out ./renders
ronana folder list | add <folder> <id…>

Every field from model get is a flag (--aspect_ratio 16:9, --seed 7). Repeat a flag to fill a list. --image, --video and --audio fill the model's media field, and accept a local file, an https URL, or a Ronana id. Local files upload first. --input '{...}' (or --input @body.json) sets the whole body for nested fields. --wait prints the result link; --out also downloads it. Add --json to any command for scripts. Run ronana --help for the full list.

generate create spends credits. Each run sends its own idempotency key, so the same command run twice makes two generations. Pass --idempotency-key to make a retry bill once.

Use Ronana from an AI app

Most AI apps connect to Ronana's remote MCP server directly. Add this URL as a remote MCP server or custom connector, then sign in when the app asks:

https://ronana.ai/mcp

Open Account > Connections on ronana.ai to see or disconnect an app.

Local MCP server

Use ronana-mcp for an app that only runs local MCP servers, or to let your AI upload files from your computer. Sign in once with ronana auth login, then add it to the app's MCP settings:

{
  "mcpServers": {
    "ronana": {
      "command": "npx",
      "args": ["-y", "-p", "@ronana/mcp", "ronana-mcp"],
      "env": { "RONANA_MCP_UPLOAD_ROOT": "/path/to/your/media" }
    }
  }
}

On Windows, start it through cmd, or the app cannot find npx:

{
  "mcpServers": {
    "ronana": {
      "command": "cmd",
      "args": ["/c", "npx", "-y", "-p", "@ronana/mcp", "ronana-mcp"],
      "env": { "RONANA_MCP_UPLOAD_ROOT": "C:\\Users\\you\\Pictures\\ronana" }
    }
  }
}

In Claude Code: claude mcp add ronana -- cmd /c npx -y -p @ronana/mcp ronana-mcp. In PowerShell, quote the separator ('--'), or PowerShell drops it.

The local server adds one tool, upload_local_reference, which uploads a file from inside RONANA_MCP_UPLOAD_ROOT. Set it to the folder that holds your media. The tool is off until it is set, and it refuses a home directory or a drive root. Files larger than 3 MB stream straight to storage, and Ronana checks each file before it adds it to your library.

Tools

A generate call takes a model and an input, exactly as get_model shows it in input_schema. A media field takes a public https URL or a Ronana id (ref_…, img_…, gen_…). Generation tools return an id immediately. Call get_generation with wait_seconds until the status is completed or failed. Apps that support MCP Apps show the image or video right in the chat.

{ "model": "google/nano-banana-2/edit", "input": { "prompt": "a lantern shop at dusk", "images": ["https://example.com/ref.png"] } }

Generations and uploads appear in your Ronana library. Your plan's storage limits and retention apply.