How to Use NanoBanana MCP in Claude Code for Terminal-Based Image Workflows

How to Use NanoBanana MCP in Claude Code for Terminal-Based Image Workflows

When you are building a product, a docs site, or a small internal tool, image work often becomes a context-switching tax: open a separate web app, upload references, write prompts, download results, and then return to the terminal. The NanoBanana MCP integration for Claude Code is useful because it moves that loop into the same place where you already plan, code, and review changes.

What you can do

The NanoBanana MCP server lets Claude Code call Gemini-powered image tools from a terminal session. The documented tool surface is intentionally small:

  • nanobanana_generate_image for text-to-image generation.
  • nanobanana_edit_image for image editing, including multi-image input.

The practical difference from a basic image generator is image understanding. You can describe a relationship between multiple inputs, such as taking an object from one image and placing it into another scene while matching angle and lighting. That makes it useful for product mockups, page illustrations, screenshot cleanup, and quick visual experiments during development.

How it works

Claude Code connects to the NanoBanana MCP server over HTTP. You register the server once with claude mcp add, pass an authorization header, and then ask Claude Code to generate or edit images in natural language. The MCP server exposes the image tools; Claude Code decides when to call them based on your request.

The documented server URL is https://nanobanana.mcp.acedata.cloud/mcp. Configuration uses an AceData Cloud API token in the HTTP header:

claude mcp add nanobanana --transport http https://nanobanana.mcp.acedata.cloud/mcp   -H "Authorization: Bearer YOUR_TOKEN"

One detail worth remembering: the header flag must be uppercase -H. Lowercase -h is interpreted as help, not as an HTTP header.

Choose the right Claude Code scope

By default, if you omit -s, Claude Code uses the local scope. That means the MCP configuration only applies to the directory where you ran the command. For a one-off project, that is fine. For everyday usage, you may prefer a wider scope.

  • local: no -s or -s local; stored in ~/.claude.json; effective only in the current project directory.
  • user: -s user; stored in ~/.claude.json; available across projects.
  • project: -s project; stored in .mcp.json in the project root; can be shared with a team.

If you use -s project, do not commit a real token to a public repository. Use an environment-variable placeholder in the shared config and keep the secret outside git.

Verify the connection before asking for images

After adding the server, run:

claude mcp list

You should see nanobanana marked as ✓ Connected. This small check saves time because image requests will fail later if the server was not registered correctly or the header was malformed.

Use case 1: multi-image composition

A common builder workflow is compositing a product, icon, device, or object into another scene. The documented example is to put the phone from a first image onto the desk in a second image, adjusting angle and lighting so the result looks natural.

In Claude Code, you can write a plain instruction like:

Put the phone from the first image onto the desk in the second image, adjusting angle and lighting to make it look natural.

Because nanobanana_edit_image supports multiple image URLs as input, you can keep the task conversational while still using the underlying image-editing tool.

Use case 2: cleanup for screenshots and docs

Another practical use is removing unwanted text from an image. For example, when a screenshot contains a watermark or a distracting label, you can ask:

Edit this image to remove the text watermark in the bottom right corner, filling the background.

This is especially handy for documentation and landing pages, where a small visual distraction can make a tutorial harder to read.

Use case 3: generate illustrations while coding

For interface states such as empty pages, error screens, or 404 pages, you can generate a lightweight illustration without leaving the terminal:

Generate an illustration of a small dinosaur crouching next to a crack in the ground looking down, for use on a 404 page.

The point is not to replace a design process. It is to reduce the cost of creating a first visual draft while you are still shaping the product experience.

A small workflow I would actually use

  1. Add the MCP server with claude mcp add.
  2. Run claude mcp list and confirm ✓ Connected.
  3. Keep reference images in a project folder or use stable image URLs.
  4. Ask Claude Code for a focused edit, not a vague redesign.
  5. Review the generated result, then iterate on one variable: composition, lighting, object placement, or text removal.

This keeps image generation close to the real development loop. You can update a docs page, generate a matching illustration, and keep the surrounding code review context in one conversation.

If you want the exact setup notes and related examples, read the original Ace Data Cloud document: Claude Code integration with NanoBanana MCP.

Comments

Post a Comment

Popular posts from this blog

Artistic QR Code API Integration Guidance

Image
Artistic QR code is an innovative technological product that combines QR codes with beautiful background images, creating practical and visually appealing works. They not only possess the functionality of traditional QR codes, allowing quick scanning and recognition by smart devices, but also incorporate artistic elements to enhance visual appeal and brand recognition. Some artistic QR codes are even generated by artificial intelligence, fully leveraging modern technology to demonstrate unparalleled innovation and uniqueness. This makes artistic QR codes widely applicable in areas such as brand marketing and advertising. In simple terms, artistic QR codes are the perfect combination of QR code scanning and artistic aesthetics. They not only provide the function of information transmission, but also enhance the user's visual experience, making each scan a delightful artistic enjoyment. Overview of Artworks Let's take a look at a few QR code works: How are they? These are ar...
Read more
Image
Nexior is an open-source project on GitHub that allows us to deploy our AI application site with one click, including AI Q&A, Midjourney drawing, knowledge base Q&A, artistic QR codes, and more. There is no need to develop AI systems, purchase AI accounts, worry about API support, or configure payment systems. It comes with zero startup costs and allows you to earn income through AI without risk. This article will introduce the deployment process of the Nexior project on Vercel. Without any programming skills, you can deploy your AI site in a few minutes and easily generate revenue. Preparation First, open the Nexior GitHub repository at: https://github.com/AceDataCloud/Nexior , then register or log in to your GitHub account. Click Fork to clone the code to your local repository, as shown below: After completing the Fork, we will get our personal repository, as shown below: In this example, the account is Germey, so we can see that we have Forked to the Germey user, wit...
Read more

A Small Tip for Using AI Agents: Don’t Ask for the Plan Too Soon

Image
A Small Tip for Using AI Agents: Don’t Ask for the Plan Too Soon Recently, while using AI agents to think through product and technical designs, I found a small but useful habit. It is not a complicated prompt trick. It is much simpler than that. Instead of asking the agent to immediately produce a plan, I first ask it to understand my requirement. If anything is unclear, I ask it to question me before writing the final proposal. In other words: Don’t generate the solution yet. First understand what I want. If anything is unclear, ask me questions. After we clarify the details, then give me the plan. This sounds ordinary, but it has improved the quality of the output quite a lot. 1. The problem with asking for a plan too early I used to work with agents in a very direct way. I would describe a rough idea and say something like: Help me design a solution. The agent would usually respond quickly with a complete-looking proposal. The structure was often clean. The logic lo...
Read more