Alibaba's Wan Ships a CLI So AI Agents Can Generate Videos Autonomously

Alibaba's Wan just shipped a CLI that turns its AI video and image generation platform into a native tool for AI agents. Instead of opening a browser, you paste one install command into your agent and it can generate images and videos on demand, all billed against the same credit balance you already have on the platform.

What changed

Until now, Wan was primarily a web platform. You'd log in, prompt, wait, download. The new CLI bridges that workflow into agentic pipelines. The Wan-skills repo, maintained by the Wan-Video team, is a suite of skills designed for AI agents, empowering them with generative capabilities through API calls. The CLI announced today is the official, first-party version of that idea, built directly into the create.wan.video platform.

The install flow is deliberately minimal: log in to the platform, find the CLI section in the bottom-left corner, copy the command, and paste it into your agent. Credits consumed through the CLI are fully synced with your existing Wan account, so there's no separate billing to manage.

What Wan actually does

Wan is an AI creative platform developed by Alibaba's Tongyi Lab that generates high-fidelity videos and images from text prompts, images, audio clips, and reference videos. The platform supports text-to-video, image-to-video, speech-to-video, and reference-to-video generation with native audio synchronization.

The model family has moved fast. The current generation covers text-to-video and image-to-video with synced audio at 720p and 1080p, plus reference-to-video, video-extend, image-edit, and text-to-image. Unlike a text-to-video model that begins only with a written description, the reference-to-video mode accepts reference media that can guide a subject's appearance, identity, movement, voice characteristics, or role in a newly generated scene.

Why a CLI matters for agents

The real unlock here is composability. An agent that can call Wan as a tool can drop video generation into any multi-step workflow: generate a product image, animate it, extend the clip, and hand back a URL, all without human intervention. This is the same pattern that's been playing out across the agentic tooling space.

The trend is real. The release of CLI tools for creative AI marks a shift from "coding agents" into "creative agents." Previous CLI tools primarily focused on code generation and development workflows. Bringing image and video generation into the agent ecosystem via CLI signals a trend where the terminal is moving beyond development and into content creation.

Wan isn't alone here. ByteDance's Jimeng, third-party wrappers like WanMCP, and platforms like Vivideo have all been building similar agent-native video tooling. What's different today is that this is the official, first-party CLI from Alibaba themselves, meaning it's the most likely to stay in sync with new model releases and platform features.

How to use it

1. Log in to create.wan.video
2. Find the CLI section in the bottom-left corner of the dashboard
3. Copy the install command shown there
4. Paste it into your agent (Claude Code, Cursor, or any compatible environment)

For teams that want to go deeper, the Wan-skills GitHub repo provides Python scripts for async task submission, status polling, OSS file upload, and resolution parsing. The underlying API uses an async pattern: you submit a job, get back a task ID, and poll until the video URL is ready.

# Conceptual async pattern used by the Wan API
import dashscope
from dashscope.video import VideoSynthesis
rsp = VideoSynthesis.async_call(
    api_key=api_key,
    model="wan2.7-i2v-2026-04-25",
    media=[{"type": "first_frame", "url": "https://..."}],
    resolution="720P",
    duration=10,
    prompt="A cyberpunk city at night, neon reflections on wet pavement"
)
# Poll rsp.output.task_id until task_status == "SUCCEEDED"
# Then read rsp.output.video_url

Pricing

Wan uses a freemium model: a usable free tier with optional paid upgrades. On the official Wan platform, credit packages run from $1.50 for 30 credits up to $100 for 3,900 credits, with membership plans starting at $5/month (billed yearly) for 300 credits per month. CLI usage draws from the same credit pool, so whatever plan you're already on applies directly.

For raw API access through Alibaba Cloud's DashScope, global rates are $0.086 per second at 720p and $0.143 per second at 1080p. A 10-second 720p clip works out to roughly $0.86 per generation at those rates.

What it's good for

• Automated content pipelines , generate social media clips, product demos, or explainer videos as part of a larger agent workflow without manual steps
• Reference-to-video workflows , the model supports single-role and multi-role video generation, multi-shot narratives, and audio-video synchronization, making it useful for character-consistent ad creative
• Image-to-video animation , Wan ships text-to-video, image-to-video, reference-to-video, and video editing modes; the image-to-video mode animates a still into a 2–15 second clip at 720p or 1080p while keeping character and product consistent
• PPT and document enrichment , the Wan-skills repo already ships a skill that converts documents or topic descriptions into PowerPoint slides with AI-generated images baked in

The CLI is available now. If you're already using Wan through the web interface, the upgrade path is a single paste into your agent's config.