NotifKit vs Novu

Novu is an open-source notification system built around a web portal and a microservices stack that needs MongoDB. NotifKit is a lean notification engine on PostgreSQL + Redis, for developers who want code-first workflows, a small infrastructure footprint and native AI integration.

The short version

Novu suits organisations that want a full web dashboard where product and marketing managers configure workflows visually. Self-hosting it means running a heavy microservices deployment (API, Worker, Web Dashboard, Ingestion, MongoDB cluster, Redis, WebSocket gateway).

NotifKit is a minimal, fast engine for developers, running on the stack they already have: PostgreSQL + Redis. Workflows live in code, versioned in Git, with under 150MB idle RAM and a native Model Context Protocol (MCP) server for AI coding assistants.

Feature comparison

Pricing & total cost of ownership (TCO)

Teams often overlook two costs: database/cluster operational maintenance and underlying transport deliverability (Twilio SMS, Resend/SES Email, FCM Push).

Total cost of ownership by send volume

Monthly TCO by volume, assuming a standard channel mix (10% SMS @ $0.0079 via Twilio, 70% Email @ $0.0001–$0.0008 via Resend/SES, 20% Push @ $0 via FCM), Novu Cloud's Pro tier ($30/mo base for 30k runs + $1.20/1k overage) at 50k sends, and Team tier ($250/mo including 250k events + $0.001/msg overage) above that:

* Self-hosting both? Novu Community Edition's licence is $0, but its 6 NestJS microservices and MongoDB need an 8GB–16GB RAM server (~$48–$150/mo in VPS hosting). NotifKit runs in <150MB idle RAM on your existing PostgreSQL with a $10–$20/mo worker, saving hundreds in memory and database complexity.

Key differences deep-dive

1. Database alignment: PostgreSQL vs MongoDB

Most web backends already run PostgreSQL. Novu means provisioning, monitoring, backing up and securing a separate MongoDB cluster just for notification state.

NotifKit uses PostgreSQL as its system of record: logs, preferences, contacts and workflow state, with standard schema migrations. Queueing and worker coordination happen in Redis Streams.

2. Native Redis Streams execution vs external bridge endpoints

Novu Framework uses an external HTTP bridge: your app exposes a serve() endpoint, and Novu's cloud/backend calls it back on every workflow step.

NotifKit runs workflows inside your infrastructure on Redis Streams and PostgreSQL state machines: no bridge round-trips, tunnels or webhook callback vulnerabilities. Workflows survive process restarts and node redeploys automatically:

import { workflow } from "notifkit";

workflow("security-alert", async ({ step, event }) => {
  const { ip, device } = event;

  // 1. Send push alert immediately
  await step.notify({
    template: "new-login-push",
    channels: ["push"],
    data: { ip, device },
  });

  // 2. Wait 10 minutes for user confirmation
  const acknowledged = await step.waitForEvent("login.confirmed", {
    timeout: "10m",
    match: { userId: event.user.id },
  });

  // 3. If unacknowledged, escalate to SMS and notify security Slack channel
  if (!acknowledged) {
    await step.notify({
      template: "urgent-login-sms",
      channels: ["sms"],
      priority: "high",
    });

    await step.notify({
      user: "usr_security_ops",
      template: "suspicious-activity",
      channels: ["slack"],
      data: { userId: event.user.id, ip, device },
    });
  }
});

3. Built for AI coding assistants via Model Context Protocol (MCP)

NotifKit ships a Model Context Protocol (MCP) server. AI agents (Claude, Cursor, Copilot, Antigravity) can check system health, trigger test workflows, preview templates and triage dead letters from your IDE, without a web dashboard.

When to choose what