AI Agent 快速開始
為 Claude Code、OpenClaw 等 AI Agent 提供 mopng.cn API 接入指南
1
取得 API Key
請先登入帳號以取得 API Key
2
安裝 MoPNG API Skill
OpenClaw / Claude Code / Cursor 可依下面三步操作;每步皆可單獨複製。
1. 安裝
在對應客戶端中執行(入口因產品而異,以各產品文件為準)。
/skill add https://gitee.com/jkin8010/mopng-api/raw/main/SKILL.md
前置依賴:uv、python3(建議 3.10+)。說明見 SKILL.md(GitHub)、Gitee
2. 設定環境變數
建立 API Key 後,此處會填入你的金鑰用於複製。
Configure environment variables (read by the skill scripts, same value as API Key)
Tell your assistant to set it globally, for example:
> Set MOPNG_API_KEY='ak_xxxxxxxxxxxxxxxx' in the global environment
If your client supports JSON env config (e.g. OpenClaw), use:
{
"env": {
"MOPNG_API_KEY": "ak_xxxxxxxxxxxxxxxx"
}
}3. 使用範例
After installation, you can use CLI-style commands in chat (relative paths are supported): - Background removal: remove-bg ./photo.jpg remove-bg ./photo.jpg --output ./result.png --async-mode - HD upscale: upscale ./photo.jpg --scale 2 upscale ./photo.jpg --scale 4 --tile-size 192 --tile-pad 24 --async-mode - Outpainting: outpainting ./photo.jpg --direction all --expand-ratio 0.5 outpainting ./photo.jpg --direction right --expand-ratio 0.3 --best-quality - Image translation: translation ./photo.jpg --target-language en translation ./photo.jpg --target-language ja --source-language zh - Text-to-image: text-to-image --prompt "A red-beaked blue magpie standing on a branch" text-to-image --prompt "A futuristic city in cyberpunk style" --output ./cyberpunk.png --model wanx-v2.5 - Image-to-image: image-to-image --input ./photo.jpg --prompt "Turn the sky into a golden sunset" image-to-image --input ./portrait.jpg --prompt "Convert to oil painting style" --strength 0.7 --output ./portrait_oil.png - List models: list-models --type text_to_image list-models --type image_to_image Tips: for long-running tasks (remove-bg/upscale), prefer --async-mode; for image-to-image, use --strength to control change level (about 0.3–0.5 light, 0.6–0.8 medium, 0.9–1.0 strong).
3
複製完整指令給 AI Agent
以下為僅 HTTP/OpenAPI 說明預覽;「複製完整指令」會包含上方 Skill 三步全文 + 本節,便於一次性傳送給助手。
Use mopng.cn OpenAPI (HTTP, Bearer auth)
API Key: ak_xxxxxxxxxxxxxxxx
Base URL: https://mo-api.mopng.cn
Image input: for local files, upload first via POST https://mo-api.mopng.cn/api/v1/images/upload, then use data.signedUrl or data.signed_url; remote URLs can be passed directly.
Async tasks: if POST returns data.status as pending/processing, poll until completed/failed (do not rely on async_mode alone):
- remove-bg / upscale: GET https://mo-api.mopng.cn/api/v1/open/image/tasks/{taskId or task_id}
- outpainting / translation / text-to-image / image-to-image: GET https://mo-api.mopng.cn/api/v1/open/ai/generations/{jobId, taskId, or task_id}
Result image: prefer first url in data.results[]; keep compatibility with result.imageUrl / image_url. Error rule: treat as failure only when code is a number and non-zero.
HTTP endpoints:
- remove-bg: POST /api/v1/open/image/remove-bg
- upscale: POST /api/v1/open/image/upscale
- outpainting: POST /api/v1/open/image/outpainting
- translation: POST /api/v1/open/image/translation
- text-to-image: POST /api/v1/open/ai/generations (type=text_to_image)
- image-to-image: POST /api/v1/open/ai/generations (type=image_to_image)
API documentation (for agents): https://mopng.cn/agent/docs
Natural language examples (works for HTTP or installed skill):
"Please remove the background of ./photo.jpg and save as ./result.png"
"Please upscale ./image.jpg by 2x"
"Please translate Chinese text in ./poster.jpg to English"
"Please outpaint ./photo.jpg by 50% on all sides"支援的 AI 工具
智慧去背
1点/张
高畫質放大
1点/张
智慧擴圖
2点/张
圖片翻譯
2点/张
文字生圖
3点/张
圖生圖
3点/张
目前選擇: 智慧去背
Base Path: /api/v1/open
Auth: Authorization: Bearer <API_KEY>
**POST** /image/remove-bg
Image: upload local files first via POST /api/v1/images/upload; for remote images, pass image_url directly.
If data.status is pending/processing, poll GET /image/tasks/{taskId|task_id}; camelCase fields are also supported (taskId, results[].url).
程式碼範例
import requests
import sys
API_KEY = "YOUR_API_KEY"
BASE_URL = "https://mo-api.mopng.cn"
# For local images, upload first. Skip for remote URLs.
base = BASE_URL.rstrip("/").replace("/api/v1/open", "") or BASE_URL
with open("image.jpg", "rb") as f:
up = requests.post(
f"{base}/api/v1/images/upload",
headers={"Authorization": f"Bearer {API_KEY}"},
files={"file": f}
)
ud = up.json().get("data") or {}
image_url = ud.get("signedUrl") or ud.get("signed_url")
# Call API
resp = requests.post(
f"{BASE_URL}/api/v1/open/image/remove-bg",
headers={"Authorization": f"Bearer {API_KEY}"},
json={
"image_url": image_url,
"output_format": "png",
"return_mask": false,
"only_mask": false,
"async_mode": false
}
)
j = resp.json()
code = j.get("code")
if code is not None and code != 0:
print("API error:", j, file=sys.stderr)
sys.exit(1)
data = j.get("data") or {}
tid = data.get("jobId") or data.get("job_id") or data.get("taskId") or data.get("task_id")
status = data.get("status")
if status in ("pending", "processing") and tid:
import time
while status in ("pending", "processing"):
time.sleep(3)
r = requests.get(
f"{BASE_URL}/api/v1/open/image/tasks/{tid}",
headers={"Authorization": f"Bearer {API_KEY}"},
)
j2 = r.json()
c2 = j2.get("code")
if c2 is not None and c2 != 0:
print("poll error:", j2, file=sys.stderr)
sys.exit(1)
data = (j2.get("data") or {})
status = data.get("status")
if status in ("completed", "failed"):
break
results = data.get("results") or []
first = results[0] if results else {}
url = first.get("url") or first.get("imageUrl") or first.get("image_url")
if not url:
r0 = data.get("result") or {}
url = r0.get("imageUrl") or r0.get("image_url")
print({"status": status, "image_url": url, "data": data})
else:
print(j)
介面說明文件(供 Agent 閱讀)
將每個工具的 endpoint、請求本體與回傳結構整理成可直接複製給 AI Agent 的 Markdown。