# AceDataCloud/MCPNanoBanana [Health: Active]

**Category:** 🎨 Art & Culture  
**Repository:** https://github.com/AceDataCloud/MCPNanoBanana  
**GitHub Stars:** 7  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/acedatacloud-mcpnanobanana

## Description
NanoBanana AI image generation and editing with virtual try-on and product placement in realistic scenes.

## Tools
Capabilities this server exposes over MCP:

- **nanobanana_generate_image** — Generate an AI image from a text prompt using Google's Nano Banana model.
- **nanobanana_edit_image** — Edit or combine images using AI based on a text prompt.
- **nanobanana_get_task** — Query the status and result of an image generation or edit task.
- **nanobanana_get_tasks_batch** — Query multiple image generation/edit tasks at once.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mcpnanobanana": {
    "command": "uvx",
    "args": ["mcp-nanobanana-pro"],
    "env": {
      "ACEDATACLOUD_API_TOKEN": ""
    }
  }
}
```

**Requires environment variables:** `ACEDATACLOUD_API_TOKEN` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What AceDataCloud/MCPNanoBanana MCP server does

AceDataCloud/MCPNanoBanana MCP server connects MCP-compatible clients to Google's Nano Banana model through the AceDataCloud API. It supports generating an image from a text prompt, editing an existing image, and combining multiple images according to a prompt. The README also identifies virtual try-on and product placement as supported use cases.

The server is intended for conversational or agent-driven image workflows. For example, an agent can create an image of a described scene, place a clothing item on a person using image URLs, or put a product into a realistic environment. Prompt quality matters: the documentation recommends describing the subject, atmosphere, lighting, camera or lens style, and visual quality characteristics.

## How it works

Requests are exposed as MCP tools. Image generation and editing return tasks that can be checked later, rather than requiring the client to perform the underlying image operation directly. `nanobanana_get_task` checks one task, while `nanobanana_get_tasks_batch` checks multiple generation or editing tasks at once.

The project offers a hosted endpoint at `https://nanobanana.mcp.acedata.cloud/mcp`. Hosted requests use a Bearer token from AceDataCloud. Claude.ai can connect through an OAuth flow, while the documented configurations for Claude Desktop, Cursor, Windsurf, Cline, and other clients use the endpoint with an authorization header.

## Setup and configuration

AceDataCloud/MCPNanoBanana MCP server can be used without local installation through the hosted endpoint. Create an account on the AceDataCloud Platform, obtain an API token from the Nano Banana API documentation page, and provide that token to the MCP client.

For local use, install the `mcp-nanobanana-pro` package from PyPI or run it with `uvx`. The local process reads the token from `ACEDATACLOUD_API_TOKEN` and starts in stdio mode by default. HTTP mode is available with `--transport http --port 8000`. The README also documents a Docker image at `ghcr.io/acedatacloud/mcp-nanobanana-pro:latest` for self-hosting.

A local Claude Desktop configuration uses `uvx` with the package name in `args` and passes the API token through the environment. Hosted configurations instead specify the streamable HTTP URL and an `Authorization: Bearer` header.

## Tools and capabilities

The four documented tools are:

- `nanobanana_generate_image`: Creates an image from a text prompt using Nano Banana.
- `nanobanana_edit_image`: Edits or combines images based on a text prompt.
- `nanobanana_get_task`: Retrieves the status and result of one image task.
- `nanobanana_get_tasks_batch`: Retrieves statuses and results for multiple image tasks.

These capabilities cover text-to-image creation, image modification, image compositing, virtual try-on, and product-scene composition when the required image inputs are supplied.

## Limitations and notes

The server depends on the AceDataCloud API and requires an AceDataCloud token for authenticated requests. The provided material does not specify API pricing, image limits, supported output formats, task-duration guarantees, or model parameter details. The hosted health-check endpoint does not require authentication, but MCP operations do.

The documentation describes image inputs as image URLs in its virtual try-on example. It does not establish that local file paths or binary uploads are accepted. Client compatibility beyond the explicitly documented configurations should be verified against the relevant MCP client.

_Full upstream README: https://allmcps.com/mcp/acedatacloud-mcpnanobanana/readme_