Checkpoint TypeScript SDK

The control layer for AI agents. Check any agent action before it runs, evaluate it against your policy, and decide: allow, stop, or escalate. Every decision is written to a tamper-evident audit log.

Install

npm install checkpoint-sdk

Quickstart

import { check, setPolicy, Policy, rule } from "checkpoint-sdk";

setPolicy(
  new Policy([
    rule("transfer_funds", { when: "amount > 25000", verdict: "escalate", to: "cfo" }),
    rule("export_records", { when: "pii_rows > 10000", verdict: "stop" }),
  ]),
);

async function processPayment(amount: number, recipient: string) {
  const verdict = await check("transfer_funds", { amount, recipient });
  if (verdict.allowed) {
    await transferFunds(amount, recipient);
  }
}

check() is async and resolves to a verdict. If the action crosses a hard limit it stops (throwing CheckpointStopped); if it needs a human it escalates.

The three verdicts

• allow: within bounds, proceeds.
• stop: crosses a hard limit, execution stops.
• escalate: needs a human, routes to the right person and waits.

Audit log

import { auditLog } from "checkpoint-sdk";

for (const e of auditLog.entries()) console.log(e.logId, e.verdict, e.action);
console.log(auditLog.verify()); // hash chain intact

Run the example and tests

npm install
npm run example
npm test

Note

This is a reference implementation with a local in-memory engine. In production the SDK forwards verdicts to the Checkpoint service and writes to a hosted audit log. The public API is identical.

License

MIT.