Run Muapi's image, video, audio, and 3D models from your terminal or connect them to a coding agent. Install the open-source CLI with npm or pip, authenticate once, and use one command shape across the catalog.

Three commands and you're shipping

  1. Install: npm install -g muapi-cli
  2. Configure: muapi auth configure
  3. Run any model: muapi run flux-dev-image -p "a cyberpunk skyline at golden hour"

Built for humans and agents

Plug into your coding agent in one step

Connect Muapi to Cursor, Windsurf, Claude Code, or Claude Desktop through the Model Context Protocol. Your agent can generate images, videos, and audio without leaving the editor.

19 tools available via MCP: generate_image, image_to_image, generate_video, image_to_video, create_music, text_to_audio, upscale_image, remove_background, face_swap, video_effects, lipsync, get_result, account_balance, list_api_keys, create_api_key, delete_api_key, upload_file, list_models, generate_3d

Install from npm: npm install -g muapi-cli; PyPI: pip install muapi-cli

MUAPI CLI & MCP

Every muapi model —from your terminal and agent

One open-source CLI to run any Muapi model: image, video, audio, and 3D. Built for terminals and coding agents such as Claude Code, Cursor, and Codex.

$npm install -g muapi-cli

For Builders

Interested in a white-label MCP?

Embed the Muapi MCP server into your product, fully branded as your own.

Three commands and you're shipping

1

Install

Node 18+ is required. One global install gives you the muapi binary everywhere.

$npm install -g muapi-cli
2

Configure

Open the access-keys page in your browser. The CLI detects a copied key and validates it live.

$muapi auth configure
3

Run any model

Use the same shape for every model: pass the model ID and inputs, receive a URL, or pipe --output-json into your scripts.

$muapi run flux-dev-image -p "a cyberpunk skyline at golden hour"

Built for humans and agents

Designed to work standalone or as a tool inside any AI coding agent.

Dynamic per-model help

Run muapi run <model> -h and the CLI fetches the live input schema, then prints the exact flags that model accepts.

Aliases that ship with your repo

Define named shortcuts in muapi.json that bundle a model and default inputs so every teammate uses the same presets.

Agent-native by design

Every command supports --output-json on stdout and progress on stderr. Add a SKILL.md so Claude Code, Cursor, and Codex know how to call it.

No magic, no daemon

The CLI never mutates your prompt or opens a background process. It calls api.muapi.ai directly and is open source under the MIT license.

What a real session looks like

From catalog browse to JSON output — three commands.

~/projects/marketing
# 1. browse the catalog
$ muapi models list --type image

# 2. inspect a model's inputs — schema fetched live
$ muapi run flux-dev-image -h
  * prompt          string
    width           number  default: 1024
    height          number  default: 1024
    output_format   string  default: "jpeg"  enum: jpeg | png | webp

# 3. run — JSON out so you can pipe it
$ muapi run flux-dev-image \
    -p "a cyberpunk skyline at golden hour" --output-json
{
  "model": "flux-dev-image",
  "outputs": ["https://cdn.muapi.ai/.../skyline.jpeg"],
  "elapsed_ms": 3241
}

Plug into your coding agent in one step

Connect Muapi to Cursor, Windsurf, Claude Code, or Claude Desktop through the Model Context Protocol. Your agent can generate images, videos, and audio without leaving the editor.

Open Cmd+Shift+P Open MCP settings and add to mcp.json:

mcp.json
{
  "mcpServers": {
    "muapi": {
      "url": "https://api.muapi.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_MUAPI_KEY"
      }
    }
  }
}

Get your key at muapi.ai/access-keys. Restart your editor after saving.

19 tools available via MCP

generate_imageimage_to_imagegenerate_videoimage_to_videocreate_musictext_to_audioupscale_imageremove_backgroundface_swapvideo_effectslipsyncget_resultaccount_balancelist_api_keyscreate_api_keydelete_api_keyupload_filelist_modelsgenerate_3d

Open source. Ready today.

MIT licensed, published on npm and PyPI, hosted on GitHub. Issues and pull requests are welcome.