Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Wan30 MCP generate a 5-second 720p video of a paper boat on a rainy street"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Generate Wan 3.0 videos from your terminal or AI assistant with your wan30.video account. MCP and CLI spend the same credit balance as the website, at the same prices. Credits are bought on the website.
Access requires remaining purchased credits. Sign-up, check-in and gift credits alone do not qualify. Buy credits at https://wan30.video/pricing/.
Node 22+, macOS or Linux (including WSL).
npm install -g @wan30/cliRelated MCP server: Wan 3.0 Prime API MCP Server
wan30 auth login
wan30 account
wan30 models
wan30 estimate --input request.json
wan30 create --input request.json --wait
wan30 download <taskId> --output result.mp4
wan30 auth logoutOn Linux, sign in with wan30 auth login --credential-store file.
request.json examples. Text to video:
{"modelId":"wan30-text-to-video","prompt":"A paper boat on a rainy street","durationSeconds":5,"resolution":"720p"}Image to video, with one HTTPS image as the first frame:
{"modelId":"wan30-image-to-video","prompt":"The crane takes off","imageUrls":["https://example.org/first-frame.jpg"]}Reference to video, with 1–10 HTTPS reference images:
{"modelId":"wan30-reference-to-video","prompt":"Move the camera slowly toward the scene","imageUrls":["https://example.org/reference.jpg"]}Every mode takes 2–30 seconds and 480p, 720p or 1080p. The default is 5
seconds at 480p. wan30 model <modelId> prints every option as a JSON Schema.
create prints a request ID on stderr. Re-running it with
--request-id <id> after a network error returns the same task and is never
charged twice. A wait timeout only stops waiting; the task keeps running.
Credentials are stored in the macOS Keychain, or with
--credential-store file in a private file (directory 0700, file 0600).
WAN30_CONFIG_DIR changes that directory. WAN30_ACCESS_TOKEN accepts a
pre-issued access token for automation. Logging out revokes this connection;
changing your website password revokes every connection.
Sign in once with wan30 auth login, then register the stdio server:
codex mcp add wan30 -- wan30 mcp
claude mcp add wan30 -- wan30 mcpTools: account_status, list_models, get_model, estimate_generation,
generate_video, list_tasks, get_task.
Clients that accept JSON configuration can launch the same stdio server with npx:
{
"mcpServers": {
"wan30": {
"command": "npx",
"args": ["-y", "@wan30/cli", "mcp"]
}
}
}The server starts and lists its tools before you sign in. A tool call made
without a sign-in tells the assistant to run wan30 auth login; no restart is
needed afterwards.
Read the account status and estimate the cost before generating. Video generation spends your website credits.
MIT. The hosted generation service is operated at wan30.video.
Available website credits and access to generation.
Inputs: no named parameters.
Quote credits before generation.
Inputs: input (required).
Spend website credits on a text, first-frame or reference video. Confirm the estimate and reuse requestId on retries.
Inputs: input (required), requestId (required).
Input options for the selected video model.
Inputs: modelId (required).
Poll your video task for its status and download URL.
Inputs: taskId (required).
Supported generation models.
Inputs: no named parameters.
Recent supported video tasks for this account.
Inputs: limit.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 7 tools
Each tool targets a distinct step: account status (credits), model discovery (list/get), cost estimation, generation, and task tracking (list/get). The list/get pairs are standard and clearly separable, and estimate_generation vs account_status are distinguished by purpose (per-request quote vs overall balance).
Six of seven tools follow a clean verb_noun convention (list_models, get_model, estimate_generation, generate_video, list_tasks, get_task). account_status breaks the pattern as a bare noun phrase, a minor deviation but still readable and consistent with the domain.
Seven tools map to a complete, well-scoped workflow (credits -> model options -> estimate -> generate -> poll). No tool feels redundant or unnecessary for the stated purpose.
The core lifecycle from credit check through generation to download URL is fully covered, including pre-flight estimation and requestId-based retry guidance. Minor gaps exist around task lifecycle management (e.g., cancel/delete a task) that agents cannot perform.
AI video, image, audio and 3D generation on 200+ models, with a credit quote before each run.
Make images and video with 30+ AI models from your assistant, on your own credits and library.
Make AI images, video, voiceovers and music, and edit video, from your assistant. Paid in credits.
Generate AI video, images and music, chat with LLMs, and manage a prepaid credits wallet.
