Introduction

Build durable AI agents and long-running tasks that survive crashes, restarts, and outages.

Nelkap runs your workflows and AI agents to completion. Every step is saved as it finishes, so a run that stops picks up right where it left off.

Your code stays in your app. When a run is due, Nelkap calls your app, and the SDK runs the workflow there.

onboardUser.ts

import { Nelkap } from '@nelkap/sdk';

import { createProfile, sendWelcomeEmail } from './activities';

const nelkap = new Nelkap();

export async function onboardUser(userId: string) {
  await nelkap.step.run('profile', () => createProfile(userId));

  await nelkap.step.sleep('1 day');

  await nelkap.step.run('welcome', () => sendWelcomeEmail(userId));
}

Primitives

Run

Run any piece of work once, and only once. Then use what it returns as many times as you want, even days later.

Sleep

Pause for a minute, a week, or a whole year. Walk away, and nothing runs until your workflow wakes up on its own.

Wait

Hold until the event that matters arrives. Then carry on with its details in hand, as if you never left.

Concurrency

Set how many runs can go at once. The rest wait their turn in order. Your app is never swamped, and no run is lost.

Scores

Score each run right in your code. Watch the scores over time, and catch a drop the day it starts.

Start here

Quickstart

Run your first workflow in a few minutes.

How it works

See how a run gets through crashes, deploys, and long pauses.