MCP Image Generator
MCP server for image generation and editing: Gemini, OpenAI and Seedream with automatic prompt enhancement
- 190 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Every request goes to the provider and is billed and API keys are paid
- Files for editing are read from disk by absolute path and sent to the cloud
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/mcp-image-generatorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Every request goes to the provider and is billed and API keys are paid
- Files for editing are read from disk by absolute path and sent to the cloud
This is third-party code. Review the repository files before installing.
What it does
mcp-image takes a short image request, turns it into a detailed prompt itself and sends it to the chosen model: Gemini (default), OpenAI GPT Image or BytePlus Seedream. The result is saved to disk and returned to the assistant as an MCP resource. Prompt enhancement keeps what you said and adds what the purpose needs: lighting and angle for a photo, layout for a poster. Text that must appear in the image stays as you wrote it in your language. Enhancement can be turned off with `SKIP_PROMPT_ENHANCEMENT=true`. Besides creation there is editing of an existing image by absolute path, quality presets `fast`, `balanced` and `quality`, aspect ratios and sizes of 1K, 2K and 4K (Seedream up to 2K). Google Search grounding is available with Gemini only. The output folder is set with `IMAGE_OUTPUT_DIR`. The provider is picked with `IMAGE_PROVIDER` together with a key.
Who it is for. Developers and designers who need illustrations, product photos and image edits straight from Codex, Cursor or Claude Code.
Good fit when
- A recipe needs a dish photo and a short request is expanded into a detailed prompt
- A poster needs text in your language that stays verbatim
- A finished picture is edited: turn the person to face right keeping the rest
- A character must look the same in the next frame and the request refers to the previous image
- The model is switched between Gemini, OpenAI and Seedream without changing how you ask
Not a fit when
- You need exact generation with no model reasoning over the request: turn enhancement off with the variable
- You have a key for none of the three providers: the server does not draw images itself
- You need a local run with no cloud requests: all three models work through an API
Example request
Generate a product photo of a ceramic mug on a wooden desk for a store listingMCP
- Transport
- stdio
- Authentication
- key is optional
Environment variables
GEMINI_API_KEYsecret- Gemini key, the default provider.
OPENAI_API_KEYsecret- OpenAI key when `IMAGE_PROVIDER=openai`.
ARK_API_KEYsecret- BytePlus ModelArk key for Seedream.
IMAGE_OUTPUT_DIR- Folder for generated images.
Limitations
Node.js 22 or newer and one provider key are needed: `GEMINI_API_KEY`, `OPENAI_API_KEY` or `ARK_API_KEY` for Seedream. Every request goes to the provider and is billed by its rules, and higher quality presets are slower and cost more. OpenAI may require organization verification before image generation. Output format is PNG or JPEG for OpenAI and Seedream and automatic for Gemini. Set an absolute `IMAGE_OUTPUT_DIR` in the MCP config: the server working directory depends on the client, otherwise images land in `./output`.
How to disable. Remove the server in the client (`claude mcp remove mcp-image`) and revoke the provider key in its dashboard; generated images stay in `IMAGE_OUTPUT_DIR`.
FAQ
How do I connect mcp-image to Claude Code?
Run `claude mcp add mcp-image --env GEMINI_API_KEY=your-api-key --env IMAGE_OUTPUT_DIR=/absolute/path/to/images -- npx -y mcp-image`. Add `--scope user` to make the server available in every project.
How do I pick the provider?
Set `IMAGE_PROVIDER` to `gemini`, `openai` or `seedream` and the matching provider key. Gemini is the default.
What does prompt enhancement do?
A short request becomes a detailed one: lighting, angle or layout is added for the purpose of the image. Text for the image stays as written. Turn it off with `SKIP_PROMPT_ENHANCEMENT=true`.
Can I edit an existing image?
Yes, pass the absolute path to the file and describe the edit, for example "turn the person to face right". The result is saved to `IMAGE_OUTPUT_DIR`.
How do I control quality?
With the `IMAGE_QUALITY` variable set to `fast`, `balanced` or `quality`, or the `quality` option on a request. Higher quality is slower and costs more.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server for generating images and video from the agent: several models, a large prompt library and parallel tasks
An MCP server for generating and editing images with Gemini: picks the model per request, renders up to 4K in any aspect ratio
RunAPI MCP server: a catalog of 160 models for images, video, music and speech, pricing, task runs and balance
A local MCP server for editing and analyzing images: the agent crops frames, finds objects and reads text from photos
The media-pipeline plugin generates images through Gemini or OpenAI gpt-image-2: a skill, CLI and MCP server install in one step.
Recraft's remote MCP server: an agent draws raster and vector images, edits a region, removes the background and upscales
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| MCP Image Generator (this card) | no | key is optional | 190 | |
| GigaChat Image MCP: Kandinsky images from Claude with no VPN | no | API key | 0 | generation through GigaChat; this one uses Gemini, OpenAI and Seedream with prompt enhancement |
| ImageSorcery MCP | no | not required | 339 | local image processing and recognition; this one creates and edits images with cloud models |
README in short
The README shows a prompt enhancement example with a comparison, a provider table and a quick start for Codex, Cursor and Claude Code. Then come edit examples, result control, environment variables and the tool reference.