digitalmakerai
v1.0.2
Published
Command line and MCP bridge for Digital Maker AI: make faceless YouTube videos, Shorts, whiteboard videos, animated podcasts, thumbnails, channel research, ebooks, planners, spreadsheets and Shopify store themes from any AI agent or terminal.
Maintainers
Readme
digitalmakerai
Digital Maker AI from the command line, and a bridge so any AI agent can use it.
Digital Maker AI makes finished faceless YouTube videos, Shorts, whiteboard videos and animated podcasts from a topic: script, scene images, voiceover, captions and music, rendered and ready to upload. It also makes thumbnails and characters, researches faceless channels and their earnings, posts to YouTube, TikTok and Instagram, and makes ebooks, digital planners, spreadsheet templates and Shopify store themes.
This package is a thin client. The list of tools and every parameter come from the server when you run it, so it never goes out of date.
Install
npm install -g digitalmakeraiNode 18 or newer. No other dependencies.
Sign in
Create a key at https://digitalmaker.ai/connect-claude (Account Settings, then the API tab or the Claude MCP tab, then Create a key), then:
digitalmakerai login dmai_live_...Or set DMAI_API_KEY in the environment. The key is a password: keep it out of scripts you share.
Use it
digitalmakerai tools # everything the server can do right now
digitalmakerai help create_video # every parameter, from the live schema
digitalmakerai account # plan and minutes left
digitalmakerai video --topic "why most people quit before their first sale" --duration_minutes 10
digitalmakerai video --orientation vertical --duration_seconds 45 --topic "3 stoic rules for a bad day"
digitalmakerai status <job id> # where it has got to, link when ready
digitalmakerai wait <job id> # poll until ready
digitalmakerai podcast --topic "the history of coffee" --minutes 8
digitalmakerai channels --niche "personal finance" --limit 10
digitalmakerai image --prompt "bold thumbnail, text WHY?" --image_type youtube_thumbnail
digitalmakerai post --item_id <video id> --platforms youtube,tiktok --title "..." --caption "..."
digitalmakerai make_planner --title "Fit Year" --niche "runners" --sections "Workout Log,Habit Tracker"
digitalmakerai check_product --product planner # download link when readyAny tool name works as a command (create_video or create-video), and the short aliases above are sugar. Lists are comma separated (--platforms youtube,tiktok), objects are JSON (--host_a '{"name":"Ava","gender":"woman"}'), and --json prints the raw result.
Every video spends minutes from the account's monthly pool, one per minute of video. A vertical Short costs a flat one minute. Video needs a Premium or Ultimate plan; images, characters, channel research, ebooks, planners and spreadsheets work on Pro (monthly allowances), and the Website Maker on every plan.
Use it from an AI agent
Claude Code
claude mcp add --transport http digitalmakerai https://digitalmaker.ai/mcp --header "Authorization: Bearer dmai_live_..."Codex CLI
codex mcp add digitalmakerai --env DMAI_API_KEY=dmai_live_... -- npx -y digitalmakerai mcpCursor, Windsurf and anything that reads an MCP JSON config
{ "mcpServers": { "digitalmakerai": { "command": "npx", "args": ["-y", "digitalmakerai", "mcp"], "env": { "DMAI_API_KEY": "dmai_live_..." } } } }OpenClaw, Hermes and other skill-based agents
Both can also connect straight to the MCP server with no CLI: see https://digitalmaker.ai/openclaw and https://digitalmaker.ai/hermes. For the skill route, install the CLI, sign in, then install the skill from ClawHub with openclaw skills install @digitalmakerai/digitalmakerai; it tells the agent which command to run for each request.
Claude and ChatGPT (the apps)
No CLI needed: add https://digitalmaker.ai/mcp as a connector. See https://digitalmaker.ai/claude and https://digitalmaker.ai/chatgpt.
Per-agent guides: https://digitalmaker.ai/agents
Notes for agents
- Call
toolsorhelp <tool>before guessing at parameters. - Starting a video spends real minutes. Confirm topic, length, orientation and style with the person first, and call once.
statusreports progress; do not poll it in a tight loop.waitpolls politely every 30 seconds.- Exit code 0 means the tool succeeded, 1 the tool refused or failed (the reason is on stdout), 2 a usage mistake (details on stderr), 3 not signed in, 4 network trouble.
