MakeMidas

透過任何 MCP 用戶端,使用 GPT Image 2、Nano Banana、Veo 3.1、Kling 和 Seedance 生成 AI 圖像與影片,共用同一個點數餘額。

託管 MCP 伺服器

npx add-mcp 'https://makemidas.com/api/mcp'

可安裝到 Claude Code、Codex、Cursor 等客戶端

文件

Connect an AI app

Add the server URL, then sign in to MakeMidas and allow access when the app asks. You can disconnect any app in Settings → Connections.

ChatGPT

In the sidebar open Plugins, then Add → Add custom MCP server. Paste the server URL, keep OAuth, accept the notice and create it, then sign in. Results show in a gallery inside the chat.

Claude

Settings → Connectors → Add custom connector, paste the server URL, then Connect and sign in. Works in Claude on the web, desktop and mobile.

Claude Code

Run this, then sign in when the browser opens:

claude mcp add --transport http makemidas https://makemidas.com/api/mcp

Cursor

Use the button, or add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "makemidas": {
      "url": "https://makemidas.com/api/mcp"
    }
  }
}

Add to Cursor

Codex CLI

Run this, then sign in when the browser opens:

codex mcp add makemidas --url https://makemidas.com/api/mcp

Gemini CLI

Run this, then sign in with /mcp auth makemidas:

gemini mcp add --transport http --scope user makemidas https://makemidas.com/api/mcp

VS Code

Run this, then sign in when the browser opens:

code --add-mcp '{"name":"makemidas","type":"http","url":"https://makemidas.com/api/mcp"}'

Or ask your coding agent to set it up:

Connect the MakeMidas MCP server (https://makemidas.com/api/mcp, Streamable HTTP, OAuth sign-in) to this app, then list the MakeMidas models to check it works.

What the assistant can do

Every generation costs credits, as on the website. Assistants are told to keep to one output on the default model unless you ask for more.

generate_imageGenerate 1–4 images from a prompt and optional reference images.
generate_videoStart a video from a prompt, a first frame, first and last frames, or references.
generate_musicStart a song or instrumental from a description, or sing your lyrics.
generate_sound_effectStart a one-shot sound effect or a seamless loop at a tempo.
generate_speechTurn text into a voiceover in one of 12 voices.
transcribe_audioTranscribe an audio or video link to text with timestamps and speakers.
compare_modelsTry one prompt on 2–4 image models side by side, each at its own price.
remove_backgroundCut the subject out of a photo as a transparent WebP.
upscale_imageUpscale an image 2x, 4x or 8x with restored detail.
upscale_videoUpscale a clip of up to 60 s to 1080P, 2K or 4K.
get_generationCheck a generation and show its result; waits up to 40 s for videos and tracks.
list_generationsShow recent images, videos and tracks.
delete_generationDelete a finished generation; its links stop working.
list_modelsModels, their options and credit prices.
get_model_paramsOne model’s options, defaults, prices and an example request.
estimate_generationPrice a request without charging, with every default filled in.
search_examplesReal results made on MakeMidas, with the model and prompt behind each.
get_creditsCredit balance and plan limits.
get_credit_activityCredit history: grants, charges and refunds, newest first.
get_credits_linkWhere to buy a plan or credit pack, with what each costs.
submit_feedbackSend a bug report, feature request or praise to the team.
save_to_notionSave a result as a page in your Notion workspace.
get_upload_linkLink to the generator for photos the app can’t pass along.

REST API

For scripts and your own apps. JSON in, JSON out; results come back as signed links.

  1. Create an API key: Settings → API keys
  2. Send a request. Images come back finished; every option except the prompt has a default.
curl https://makemidas.com/api/v1/generations \
  -H "x-api-key: $MAKEMIDAS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"kind":"image","prompt":"a lighthouse at dawn, film photo","aspectRatio":"16:9"}'

Videos return right away with status running (202). Ask for them again with wait=40 until they are done:

curl "https://makemidas.com/api/v1/generations/<id>?wait=40" \
  -H "x-api-key: $MAKEMIDAS_API_KEY"

Music, sound effects and voiceovers work like videos: send kind audio with task music, sfx or speech (for speech, the prompt is the text to speak, voice picks one of the model's voices and, on Gemini models, delivery says how to read it), then ask for the result with wait=40. Suno returns two or more takes as separate outputs.

curl https://makemidas.com/api/v1/generations \
  -H "x-api-key: $MAKEMIDAS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"kind":"audio","task":"music","model":"google/lyria-3.5","prompt":"warm lo-fi hip hop, dusty piano, vinyl crackle"}'
GET /api/v1/meThe account behind the key
GET /api/v1/creditsCredit balance and plan limits
GET /api/v1/credits/activityCredit history: grants, charges and refunds, newest first;?before=&limit=1..100
GET /api/v1/credits/purchase-linkWhere to buy a plan or credit pack, with the plans and packs on sale; creates no payment
GET /api/v1/modelsModels, their options and credit prices
GET /api/v1/models/{id}One model: its options, defaults, prices and an example request
POST /api/v1/uploadsUpload a reference image (multipart file, under 10 MB)
POST /api/v1/audio-uploadsUpload audio or video to transcribe (the file as the body, or JSON with a url); returns fileId, its length and the price
POST /api/v1/video-uploadsUpload a video to upscale (the file as the body, or JSON with a url); returns fileId, its length and the price
POST /api/v1/generationsGenerate images (201), or start a video or audio job (202)
POST /api/v1/generations/estimatePrice a request without charging; same body as creating one. To create it, send the returned request with quotedCredits set to credits
POST /api/v1/generations/compareRun one prompt on 2–4 image models at once; each is a generation at its own price, and the total must fit the balance
GET /api/v1/generationsYour generations, newest first
GET /api/v1/generations/{id}One generation;?wait=1..40 waits for a video
DELETE /api/v1/generations/{id}Delete a finished generation: it leaves your list and its links stop working (409 while it runs)
POST /api/v1/feedbackSend feedback: kind bug, feature, praise or other, a message up to 2,000 characters, and an optional generationId

Full description (OpenAPI 3.1): /openapi.json

Command line

Sign in once in the browser, then generate from your terminal or from scripts.

npm install -g @makemidas/cli
makemidas auth login
makemidas image "a lighthouse at dawn" --aspect 16:9 --out ./out

Add the agent skill so your coding agent knows the commands and options:

npx skills add makemidas-ai/skills --skill makemidas

In Claude Code you can also install it as a plugin, with the MCP server included:

/plugin marketplace add makemidas-ai/skills
/plugin install makemidas@makemidas-skills

Save to Notion

Connect a Notion workspace to save a result, with its prompt and settings, as a private Notion page: from the generator, or by asking ChatGPT or Claude to save it.

Connect Notion

Credits and limits

  • API and app requests spend the same credits as the website, at the same prices.
  • There are no free credits: every generation spends credits from a plan or a credit pack, and either unlocks every model.
  • Up to 120 requests and 10 new generations a minute per account; how many run at once depends on the plan.
  • Result links are signed and expire after an hour; ask for the generation again for fresh ones.
  • Apps sign in with OAuth and can be disconnected any time. API keys can be revoked in Settings.
  • What a connected app or API key can read and do is covered in the Privacy Policy. See plans and credit packs