Nano Banana

@mindstone/mcp-server-nano-banana

Nano Banana MCP server — Google Gemini image generation and editing via Model Context Protocol. Generate images from text descriptions and edit existing images using Google Gemini's AI capabilities.

Status

• Version: 0.5.0 · npm
• Auth: API key (GEMINI_API_KEY)
• Tools: 3 (image-generation, image-editing)
• Surface: cloud-api
• Machine-readable: STATUS.json

Requirements

• Node.js 20+
• npm

One-click install

After clicking the button, your host will prompt you to fill: GEMINI_API_KEY, NANO_BANANA_GEMINI_TIMEOUT_MS, NANO_BANANA_BRIDGE_TIMEOUT_MS.

Manual config for Claude Desktop / Claude Code / Goose / Continue.dev (Nano Banana)

{

"mcpServers": { "Nano Banana": { "command": "npx", "args": [ "-y", "@mindstone/mcp-server-nano-banana" ], "env": { "GEMINI_API_KEY": "", "NANO_BANANA_GEMINI_TIMEOUT_MS": "180000", "NANO_BANANA_BRIDGE_TIMEOUT_MS": "30000" } } } }

Quick Start

Install & build

cd <path-to-repo>/connectors/nano-banana

npm install npm run build

npx (once published)

npx -y @mindstone/mcp-server-nano-banana

Local

node dist/index.js

Configuration

Environment variables

• GEMINI_API_KEY — Google Gemini API key
• MCP_WORKSPACE_PATH — optional workspace path for saving generated images
• MCP_HOST_BRIDGE_STATE — optional path to a host bridge state file used for credential management
• MINDSTONE_REBEL_BRIDGE_STATE — backwards-compatible alias for MCP_HOST_BRIDGE_STATE
• NANO_BANANA_GEMINI_TIMEOUT_MS — optional override (positive integer ms) for the outbound Gemini request timeout. Default: 180000 (3 min). Increase this if you use gemini-3-pro-image and see TIMEOUT errors on slow generations; decrease it if you want tighter bounds.
• NANO_BANANA_BRIDGE_TIMEOUT_MS — optional override (positive integer ms) for requests to the host bridge. Default: 30000.

Host configuration examples

Claude Desktop / Cursor

{

"mcpServers": { "NanoBanana": { "command": "npx", "args": ["-y", "@mindstone/mcp-server-nano-banana"], "env": { "GEMINI_API_KEY": "your-gemini-api-key" } } } }

Local development (no npm publish needed)

{

"mcpServers": { "NanoBanana": { "command": "node", "args": [" /connectors/nano-banana/dist/index.js"], "env": { "GEMINI_API_KEY": "your-gemini-api-key" } } } }

Tools (3)

Configuration

• configure_nano_banana_api_key — Save your Gemini API key for image generation

Image generation

• nano_banana_generate — Generate images from text descriptions
• nano_banana_edit — Edit existing images using AI; accepts up to 14 reference images (source_image_paths) for multi-image composition/fusion. Each source may be a workspace file path or an https:// URL — every source is capped at 20MB per image (local files included) with a combined-size cap across sources, and remote fetches are SSRF-guarded (HTTPS only, no userinfo, private/loopback/link-local hosts refused with DNS re-validation, redirects re-validated per hop, PNG/JPEG/WebP only)

Both image tools accept an optional save_path inside the workspace. Saves never overwrite: if a file (or symlink) already exists at save_path, the tool returns a SAVE_EXISTS error (with the image still included inline) — choose a different path or delete the existing file. Saved files are written owner-only (0600).

Generated and edited output is identified from its bytes before it is returned or saved. PNG, JPEG, and WebP are supported; other or unrecognised formats (including GIF) return a structured UNSUPPORTED_IMAGE_FORMAT error instead of being relabelled.

Output resolution

Both image tools accept an optional image_size parameter ("1K", "2K", or "4K"; default "1K" ≈ 1024px). Higher resolutions are only honoured by the Gemini 3 image models (gemini-3.1-flash-image, the default, gemini-3-pro-image and gemini-nano-banana-2.1); the legacy gemini-2.5-flash-image always produces ~1K output and rejects an explicit image_size.

gemini-nano-banana-2.1 (Nano Banana 2.1) can be chosen with model; it is not the default. Google enables thinking on it by default (level medium, billed), so it is slower than gemini-3.1-flash-image. It takes 1K, 2K or 4K.

The model parameter also accepts the retired preview ids gemini-3.1-flash-image-preview and gemini-3-pro-image-preview, which Google shut down on 2026-06-25; a request naming one is sent as its generally available id (gemini-3.1-flash-image, gemini-3-pro-image).

Licence

FSL-1.1-MIT — Functional Source License, Version 1.1, with MIT future licence. The software converts to MIT licence on 2030-04-08.

Host-authorized local inputs

files/readModes version 1 is an experimental extension used by these connectors, not part of the MCP standard. A compatible host (the app running the connector) decides which input files a tool call may read. The connector opens only the host-approved physical path, after resolving symlinks. The host supplies this permission information with each call; it is not a tool argument you need to set. Missing or invalid permission information, or a file absent from the approved list, stops the read.

Clients that do not negotiate this extension keep the connector's existing standalone file restrictions. This extension changes input-read permissions; existing output destinations, upload formats, size limits and file checks apply.