Installable skillMIT

Kling AI Webhook Configuration

Configure webhook callbacks for Kling AI task completion. Use when building

BY jeremylongshoreUPDATED Oct 27, 2026
INTENDED FOR✳Claude Code>_CodexInstall in a listed agent.

instructions.md

---
name: klingai-webhook-config
description: 'Configure webhook callbacks for Kling AI task completion. Use when building
  event-driven

  pipelines or replacing polling. Trigger with phrases like ''klingai webhook'', ''kling
  ai callback'',

  ''klingai notifications'', ''video completion webhook''.

  '
allowed-tools: Read, Write, Edit, Bash(npm:*), Grep
version: 1.18.0
license: MIT
author: Jeremy Longshore <jeremy@intentsolutions.io>
tags:
- saas
- kling-ai
- webhooks
- callbacks
compatibility: Designed for Claude Code
---
# Kling AI Webhook Configuration

## Overview

Instead of polling task status, pass a `callback_url` when creating a task. Kling AI will POST the completed task result to your URL when generation finishes. This eliminates polling overhead and reduces API calls.

**Supported on:** All video generation endpoints (`text2video`, `image2video`, `video-extend`, `lip-sync`, `effects`)

## How It Works

1. Include `callback_url` in your task creation request
2. Kling queues the task normally
3. When task reaches terminal state (`succeed` or `failed`), Kling POSTs the full result to your URL
4. Your webhook handler processes the result

## Sending a Task with Callback

```python
import jwt, time, os, requests

BASE = "https://api.klingai.com/v1"

def get_headers():
    ak, sk = os.environ["KLING_ACCESS_KEY"], os.environ["KLING_SECRET_KEY"]
    token = jwt.encode(
        {"iss": ak, "exp": int(time.time()) + 1800, "nbf": int(time.time()) - 5},
        sk, algorithm="HS256", headers={"alg": "HS256", "typ": "JWT"}
    )
    return {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}

# Create task with callback
response = requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
    "model_name": "kling-v2-master",
    "prompt": "A futuristic city skyline at night with neon lights",
    "duration": "5",
    "mode": "standard",
    "callback_url": "https://your-app.com/webhooks/kling",  # your endpoint
})

task_id = response.json()["data"]["task_id"]
print(f"Task {task_id} submitted with callback -- no polling needed")
```

## Webhook Receiver (Flask)

```python
from flask import Flask, request, jsonify
import hmac
import hashlib
import json

app = Flask(__name__)

@app.route("/webhooks/kling", methods=["POST"])
def kling_webhook():
    payload = request.get_json()

    task_id = payload["data"]["task_id"]
    status = payload["data"]["task_status"]

    if status == "succeed":
        video_url = payload["data"]["task_result"]["videos"][0]["url"]
        print(f"Task {task_id} complete: {video_url}")
        # Download video, store to S3, notify user, etc.
        process_completed_video(task_id, video_url)
    elif status == "failed":
        error = payload["data"].get("task_status_msg", "Unknown error")
        print(f"Task {task_id} failed: {error}")
        handle_failure(task_id, error)

    return jsonify({"received": True}), 200
```

## Webhook Receiver (Express.js)

```javascript
import express from "express";
const app = express();
app.use(express.json());

app.post("/webhooks/kling", (req, res) => {
  const { data } = req.body;
  const { task_id, task_status } = data;

  if (task_status === "succeed") {
    const videoUrl = data.task_result.videos[0].url;
    console.log(`Task ${task_id} complete: ${videoUrl}`);
    processVideo(task_id, videoUrl);
  } else if (task_status === "failed") {
    console.error(`Task ${task_id} failed: ${data.task_status_msg}`);
  }

  res.json({ received: true });
});

app.listen(3000);
```

## Callback Payload Shape

```json
{
  "code": 0,
  "message": "success",
  "data": {
    "task_id": "abc123...",
    "task_status": "succeed",
    "task_status_msg": "",
    "task_result": {
      "videos": [{
        "id": "vid_001",
        "url": "https://cdn.klingai.com/...",
        "duration": "5.0"
      }]
    }
  }
}
```

## Webhook Reliability Pattern

```python
import time
from collections import defaultdict

class WebhookManager:
    """Track webhook delivery and fall back to polling on failure."""

    def __init__(self, poll_fallback_sec: int = 300):
        self.pending = {}  # task_id -> submission_time
        self.poll_fallback_sec = poll_fallback_sec

    def register(self, task_id: str):
        self.pending[task_id] = time.time()

    def mark_received(self, task_id: str):
        self.pending.pop(task_id, None)

    def get_stale_tasks(self) -> list:
        """Tasks that haven't received a callback within threshold."""
        now = time.time()
        return [tid for tid, submitted in self.pending.items()
                if now - submitted > self.poll_fallback_sec]

    def fallback_poll(self, client):
        """Poll stale tasks that missed their callback."""
        for task_id in self.get_stale_tasks():
            try:
                result = client._get(f"/videos/text2video/{task_id}")
                status = result["data"]["task_status"]
                if status in ("succeed", "failed"):
                    self.mark_received(task_id)
                    return result
            except Exception:
                pass
```

## Requirements for Your Webhook Endpoint

| Requirement | Detail |
|------------|--------|
| Protocol | HTTPS only |
| Response | Return `2xx` within 5 seconds |
| Availability | Must be publicly reachable |
| Idempotency | Handle duplicate deliveries gracefully |
| Timeout | Kling retries on timeout, so process async |

## Prerequisites

- An allowlisted HTTPS endpoint, secret references for signature verification, sandbox event fixture, idempotency store, redaction policy, retention window, and incident owner.

## Instructions

1. Register only a sandbox callback during testing and validate signatures before parsing or queueing an event.
2. Enforce idempotency, schema validation, least-privilege downstream routing, redacted logging, and draft-only handling for generated assets.
3. Reject, quarantine, and alert on unsigned, malformed, duplicate, unapproved-destination, policy, or retention-drift events; do not retry unsafe payloads.
4. Promote a callback route only after owner approval; retain aggregate redacted evidence and delete test events at the approved boundary.

## Output

Produce a webhook receipt with endpoint classification, event correlation ID, signature/schema/idempotency results, downstream destination, draft-only and policy checks, alert owner, retention/deletion proof, and rollback reference. Exclude payloads, prompts, asset URLs, and secrets.

## Error Handling

| Condition | Response |
|---|---|
| Signature or schema validation fails | Return a safe failure response, quarantine redacted metadata, and do not forward the event. |
| Duplicate, policy, or destination violation | Deduplicate or reject the event, remove temporary artifacts, and notify the owner. |

## Examples

`endpoint=sandbox-allowlisted; event=opaque-42; signature=pass; schema=pass; idempotency=pass; destination=draft-only; cleanup=verified` is valid canary evidence.

## Resources

- [API Reference](https://app.klingai.com/global/dev/document-api/apiReference/model/textToVideo)
- [Developer Portal](https://app.klingai.com/global/dev)
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.