Kling AI for OpenClaw

Generate images and videos with Kling AI directly from OpenClaw.

The plugin connects to Kling AI's official global MCP service through OAuth. It does not require an API key or run a local MCP server.

Features

• Text-to-image, image-to-image, editing, and controlled variations
• Text-to-video, image-to-video, motion control, and multi-shot video
• Reference uploads, Elements, and motion library access
• Account, credit, task status, and result queries
• A final parameter and credit confirmation before each generation request

Install

Requires OpenClaw 2026.9.4 or a compatible release.

openclaw plugins install clawhub:kling-ai-global-openclaw

Check the plugin:

openclaw plugins inspect kling-ai-global --json

Restart the Gateway if the running instance does not load the plugin automatically:

openclaw gateway restart

Sign in

Register the OAuth MCP server once. OpenClaw 2026.9.4 reads mcp login targets from its operator-managed MCP configuration:

openclaw mcp add kling-ai-global \
  --url https://kling.ai/mcp/plugin \
  --transport streamable-http \
  --auth oauth \
  --connect-timeout 30 \
  --timeout 60 \
  --no-probe

Then sign in:

openclaw mcp login kling-ai-global

Follow the terminal prompt to sign in and authorize Kling AI in your browser. OpenClaw stores and refreshes the OAuth credentials; the plugin does not read or store tokens.

Verify the connection:

openclaw mcp doctor kling-ai-global --probe

A healthy connection reports kling-ai-global: ok.

Use

After installation and sign-in, describe what you want in an OpenClaw conversation.

Image example:

Create a 4:5 skincare ad with Kling AI. Place a frosted white serum bottle on a light gray
stone surface, use soft side lighting, leave room for a headline, and do not add people or text.

Video example:

Create a 10-second, 16:9, 1080p coffee ad with Kling AI.
Start on a close-up of pour-over coffee and slowly push in. Use warm natural light and no captions.

When attaching an image, state how it should be used:

Use this image as the first frame of a 5-second, 9:16 video.
Keep the person, clothing, and background unchanged. Have the person look up gently as the camera pushes in.

Before submission, the plugin shows the final model, dimensions or resolution, duration, and output count, and explains that the request consumes Kling AI credits. It creates the task only after explicit confirmation.

To check an existing task:

Check task <generationId>. Do not submit it again.

Update

openclaw plugins update kling-ai-global

Restart the Gateway and open a new conversation if the updated tools do not appear.

Service region

This branch uses the Kling AI global MCP service:

https://kling.ai/mcp/plugin

Regional accounts, credits, and tasks are separate. To switch accounts:

openclaw mcp logout kling-ai-global
openclaw mcp login kling-ai-global

Troubleshooting

openclaw plugins inspect kling-ai-global --json
openclaw mcp status --verbose
openclaw mcp doctor kling-ai-global --probe

If the tool list is stale:

openclaw mcp reload

If mcp login reports No MCP server named "kling-ai-global", run the registration command in Sign in, then retry login.

Then open a new conversation. If a submission times out or returns an ambiguous result, keep the generationId or taskTraceId and query the original task instead of submitting it again.

Development

git clone https://github.com/klingai-dev/openclaw-plugin.git
cd openclaw-plugin
git switch global
npm test
npm run pack:release

Install the local archive:

openclaw plugins install --force --accept-capabilities ./dist/kling-ai-global-openclaw-1.1.17.zip

See docs/COMPATIBILITY.md for compatibility and validation details.

License

MIT