Kling AI for Pi

Create and monitor Kling AI images and videos directly from Pi, using Kling's browser-based OAuth flow instead of an API key. The package includes its own remote MCP connection, Agent Skills, and MCP Apps result rendering.

What you get

• Text-to-image, image-to-image, text-to-video, and image-to-video workflows.
• Dedicated image and video Skills for prompt construction, scene patterns, reference roles, camera motion, multi-shot continuity, and quality checks.
• Account and credit queries, uploads, task status, motion control, the motion library, and Element creation, inspection, updates, and deletion when the live server exposes them.
• Global and China Kling accounts, with exactly one region active at a time.
• Browser OAuth with credentials stored by the operating-system credential store.
• A confirmation step immediately before every credit-consuming generation.
• Task polling and MCP Apps result rendering inside Pi when the host supports it.

Requirements

• Pi with Node.js 22.19 or newer.
• A Kling account for the region you choose: Global or China.

No Kling API key is required.

Quick start

1. Install and start Pi

pi install npm:kling-ai-pi
pi

If you previously installed this repository as a local Pi package, remove that source first. pi list should show only the npm package, not both an npm entry and a local repository path.

2. Choose the account region

Run these commands inside Pi:

/kling-region
/kling-region global

global is the default for a new installation. Use /kling-region china for a China account.

3. Authorize Kling AI

Still inside Pi, run:

/mcp-auth Plugin-Pi-kling-ai
/mcp reconnect Plugin-Pi-kling-ai
/mcp

Complete the Kling sign-in and authorization in the browser. Setup is working when /mcp shows exactly one Plugin-Pi-kling-ai server connected for the selected region.

4. Verify without spending credits

Send this prompt in Pi:

Check my current Kling account and region. Do not create an image or video and do not call any generation tool.

Account and credit checks are read-only. They do not submit a generation.

Create with Kling AI

You can ask naturally. The installed kling-ai Skill handles account, cross-media, upload, Element, motion-library, and task operations, then routes creative work to the dedicated kling-ai-generate-image or kling-ai-generate-video Skill. Every Skill discovers the current remote tool schemas before choosing a model or argument.

Generate an image

Use Kling AI to create a cinematic 16:9 image of a futuristic city at blue hour. Show me the final settings and wait for my confirmation before submitting.

Generate a video

Use Kling AI to create a cinematic 5-second, 16:9, 1080p video of a paper spaceship flying through a rainy neon city. Show me the final settings and wait for my confirmation before submitting. After submission, keep checking the same task until the result is displayed.

Generate from an image

Attach or reference the local image in your Pi task, then ask:

Animate this image with subtle camera movement for 5 seconds. Preserve the composition, show me the final settings, and wait for my confirmation before submitting.

Before a billable tool call, Pi should show the final workflow, model, duration or resolution, and aspect ratio. Reply with an explicit confirmation only when those settings are correct. The package submits at most once for that approved request and does not blindly retry an ambiguous or timed-out submission.

For a normal deliverable, the Skills prefer a supported 2k image model and 1080p video model. They use 4k when supported for commercial, post-production, fine-material, or crop-heavy work, and reserve lower-quality settings for drafts or explicit speed/cost requests. The current who_am_i response remains authoritative for every model and enum.

Region and account switching

Region	MCP endpoint	Pi server key
Global (default)	https://kling.ai/mcp	Plugin-Pi-kling-ai
China	https://klingai.com/mcp	Plugin-Pi-kling-ai

Finish or record any active generation IDs before changing regions. Then run:

/mcp logout Plugin-Pi-kling-ai
/kling-region china
/mcp-auth Plugin-Pi-kling-ai
/mcp reconnect Plugin-Pi-kling-ai

Replace china with global when switching back. The package never activates both endpoints at once. Do not add either endpoint separately to .mcp.json. The packaged mcp.global.json and mcp.china.json files are internal templates, not user configuration. Do not copy or symlink them; /kling-region is the only authoritative way to inspect the active region.

Troubleshooting

EPERM: process.cwd or uv_cwd

The terminal's current directory is unavailable or blocked by macOS. Start from an accessible directory and retry:

cd ~
pi list

A local package is still installed

Check the source:

pi list

Remove the old local source, then install the npm package:

pi remove /absolute/path/to/pi-plugin
pi install npm:kling-ai-pi

OAuth or tools are unavailable

Inside Pi, run:

/kling-region
/mcp-auth Plugin-Pi-kling-ai
/mcp reconnect Plugin-Pi-kling-ai
/mcp

Do not paste Kling tokens, cookies, API keys, or authorization headers into chat or configuration files. On a headless machine, follow the adapter's redirect-URL paste flow.

A submission timed out

Do not submit it again immediately. Ask Pi to query existing Kling tasks and preserve the original generationId or taskTraceId so a delayed acceptance cannot be charged twice.

Local development

For a local checkout or extracted release archive:

cd /absolute/path/to/pi-plugin
npm install --omit=dev --ignore-scripts
npm run check
npm test
pi install /absolute/path/to/pi-plugin

Pi loads local packages in place, so the dependency install is required for a local directory. npm-published packages install their runtime dependencies through Pi's package manager.

Distribution

• npm: kling-ai-pi
• Pi Package Gallery: kling-ai-pi
• Release checks: docs/RELEASE_CHECKLIST.md

This package contains only the Pi extension, its Skills, regional remote MCP configuration, and release checks. It does not bundle a local MCP server or the repository's development-only mcp-app/.