Mocai CLI

Turn ordinary video into production-ready motion capture from your terminal or your AI agent. Mocai reconstructs the people in a clip and gives you rigged animation as FBX, GLB, USD or BVH, retargeted onto Mixamo, UE5 Manny or MetaHuman if you like.

The same tool ships on npm and PyPI. Use whichever runtime you have:

npx @mocai/cli process clip.mp4       # Node 18+
uvx mocai process clip.mp4            # Python 3.9+ (or: pipx install mocai)

Sign in

mocai login

This opens your browser to sign in to Mocai (a free account works) and stores the sign-in in ~/.config/mocai/credentials.json (%APPDATA%\mocai on Windows). Both the npm and the PyPI versions share it. If your browser can't reach the CLI after you click Allow (WSL, SSH, containers), copy the address it landed on and paste it into the terminal. For scripts and CI, set MOCAI_API_KEY to a key from mocai.ai/app/developers instead.

Process videos

mocai process clip.mp4                          # FBX into ./mocai-out
mocai process takes/ --exports fbx,glb --rigs ue5 --out mocap
mocai process a.mp4 b.mov --kinds keypoints-json --json

Give files or folders (every video inside is processed). Videos upload in parallel, process on Mocai's GPUs, and the exports land in --out (default ./mocai-out).

| Option | Values | | --- | --- | | --exports | fbx, glb, usd, bvh (default fbx) | | --rigs | mixamo, ue5, metahuman: also retarget and download fbx-<rig> | | --kinds | usd-mesh, keypoints-json, keypoints-csv | | --max-persons N | performers to capture (default: your plan's cap) | | --smoothing | off, standard, strong | | --root-motion | drop-in (default) or performer | | --reference-pose | none (default), t-pose or a-pose: a reference-pose frame before the motion, for manual retargeting | | --no-hands | skip hand and finger tracking | | --no-mesh | skeleton only, no skinned performer mesh | | --json | machine-readable result on stdout |

Other commands: mocai account (plan and usage), mocai takes, mocai download <takeId> --kinds fbx,glb, mocai logout.

Use it from an AI agent (local MCP server)

mocai mcp runs a local MCP server, so Claude Desktop, Claude Code, Cursor and other MCP clients can process the videos on your machine with one tool call (process_videos). Point it at files or whole folders (~/shoot works); the exports land in a mocai-out folder next to the videos. Add it to your client's MCP config:

{
  "mcpServers": {
    "mocai": { "command": "npx", "args": ["-y", "@mocai/cli", "mcp"] }
  }
}

or with Python: "command": "uvx", "args": ["mocai", "mcp"]. Claude Code: claude mcp add mocai -- npx -y @mocai/cli mcp. Claude Desktop: download the one-click extension from mocai.ai/mocai.mcpb and open it. The first tool call opens your browser to sign in.

Prefer a hosted connector instead (claude.ai, ChatGPT, Gemini, Grok)? Add https://api.mocai.ai/v1/mcp as a custom connector. The Agents page has step-by-step setup for each app.

Exit codes

0 success, 1 a video or request failed, 2 usage error, 3 not signed in.