1

Install the SDK

pip install cheqpoint
2

Get your Connection Key

Go to Settings → Connection Keys in your Cheqpoint dashboard. Create a new key and copy it.

CHEQPOINT_API_KEY=ck_live_xxxxxxxxxxxx
3

Send your first request

Wrap any AI agent action with a single function call. The SDK blocks until a human decides (or the request times out).

Python

python
# Works as a drop-in for LangGraph interrupt()
# Works with Intercom webhooks
# Works with Zendesk triggers
# Works with any Python AI agent

import cheqpoint

cheqpoint.configure(api_key="ck_live_xxxxxxxxxxxx")

result = cheqpoint.request(
    action="process_refund",
    payload={"customer_id": "cus_123", "amount": 150},
    context="Customer requested refund for order #8821",
)

if result.approved:
    process_refund(result.payload)
else:
    print(f"Declined: {result.reason}")

Node.js

javascript
// Works as a drop-in for LangGraph interrupt()
// Works with Intercom webhooks
// Works with Zendesk triggers
// Works with any Python/Node.js AI agent

import { Cheqpoint } from "@cheqpoint/sdk";

const cheq = new Cheqpoint({ apiKey: process.env.CHEQPOINT_API_KEY });

const result = await cheq.request({
  action: "process_refund",
  payload: { customerId: "cus_123", amount: 150 },
  context: "Customer requested refund for order #8821",
});

if (result.approved) {
  await processRefund(result.payload);
} else {
  console.log(`Declined: ${result.reason}`);
}
4

Review in your inbox

Open Review Requests in your dashboard. You'll see the request with full context. Approve, decline, or modify - the SDK receives the decision instantly.

5

Done

That's it. Your AI agent now has a human checkpoint. Every decision is automatically logged to your audit trail.

Open dashboard →LangGraph migration →