Fal.ai OpenAI Image MCP Server
# Fal.ai OpenAI Image MCPサーバー
fal.aiのGPT image-1 text-to-image APIを使用して画像を生成するFastMCPサーバーです。
## セットアップ
1. 依存関係をインストール:
```bash
npm install
```
2. 環境変数としてAPIキーを設定:
```bash
export FAL_KEY="your_fal_api_key_here"
export OPENAI_API_KEY="your_openai_api_key_here"
```
3. プロジェクトをビルド:
```bash
npm run build
```
4. サーバーを実行:
```bash
npm start
```
## 使用方法
サーバーは以下のツールを提供します:
### generate_image
テキストプロンプトを使用して画像を生成します。
**パラメーター:**
- `prompt` (必須): 生成する画像のテキスト説明
- `image_size` (オプション): 生成する画像のサイズ ("auto", "1024x1024", "1536x1024", "1024x1536")
- `num_images` (オプション): 生成する画像の数 (1-4)
- `quality` (オプション): 生成する画像の品質 ("auto", "low", "medium", "high")
- `background` (オプション): 背景のタイプ ("auto", "transparent", "opaque")
**例:**
```json
{
"prompt": "夕暮れの静謐なサイバーパンク都市風景",
"image_size": "1024x1024",
"num_images": 1,
"quality": "high"
}
```
## Claude Codeでの使用
Claude Codeでこのサーバーを使用するには、以下のJSONをMCP設定に追加してください:
```json
{
"mcpServers": {
"fal-openai-image": {
"command": "node",
"args": ["/path/to/fal_openai_image1/dist/index.js"],
"env": {
"FAL_KEY": "your_fal_api_key_here",
"OPENAI_API_KEY": "your_openai_api_key_here"
}
}
}
}
```
または、開発モードで実行する場合:
```json
{
"mcpServers": {
"fal-openai-image": {
"command": "npm",
"args": ["run", "dev"],
"cwd": "/path/to/fal_openai_image1",
"env": {
"FAL_KEY": "your_fal_api_key_here",
"OPENAI_API_KEY": "your_openai_api_key_here"
}
}
}
}
```
## 環境変数
- `FAL_KEY`: 必須。fal.ai API認証用のAPIキー
- `OPENAI_API_KEY`: 必須。BYOK (Bring Your Own Key) システム用のOpenAI APIキーTDQS
Scored across 1 tool
With only one tool, there is no possibility of ambiguity or overlap between tools. The tool 'generate_image' has a clearly defined and distinct purpose for generating images via a specific API.
A single tool inherently has perfect naming consistency as there are no other tools to compare against. The name 'generate_image' follows a clear verb_noun pattern, which would be consistent if more tools were added.
A single tool is too few for a server with a domain like image generation, which could reasonably include operations like listing models, editing images, or fetching image details. This minimal set may cause agents to hit dead ends or lack flexibility in handling related tasks.
The server is severely incomplete for an image generation domain, covering only text-to-image generation. Obvious gaps include no tools for model management, image editing, retrieval of generated images, or error handling, which limits agent workflows and could lead to failures in broader tasks.