Blog
COURIERNOTIFICATIONS LANDSCAPEENGINEERING

Vibe Coding Notifications: How to Use Courier with Cursor or Claude Code

Kyle Seyler

January 22, 2026

a guide for ai assisted development: Notification infrastructure

Table of contents

The Problem with AI-Generated Notification Code

Step 1: Connect Courier MCP to Cursor

Step 2: Validate Your Setup with the CLI

Step 3: Prompt with Context

What You Can Build

When This Approach Works Well

Comparing Notification Platforms for AI-Assisted Development

Resources

Getting Started

Vibe Coding Notifications: How to Use Courier with Cursor or Claude Code

Read Time: 8 Minutes

Building apps with Cursor, Claude Code, or Codex changes how quickly you can ship features. UI components, API integrations, database schemas—AI coding tools handle these faster than ever. But when you prompt "add multi-channel notifications to my app," something interesting happens.

The AI generates basic send functions quickly. Then it starts scaffolding routing logic, fallback handling, delivery tracking, template management. The generated code keeps growing. You're suddenly reviewing hundreds of lines of infrastructure you didn't plan to maintain.

This guide covers a better approach: connecting Cursor to Courier's MCP server so it generates working notification code against your actual account, not generic examples.


The Problem with AI-Generated Notification Code

Without context about your specific setup, AI coding tools default to building notification infrastructure from scratch. They don't know what templates you've already created, what providers you've configured, or how your user data is structured.

The result is generic code that needs adaptation. You end up debugging generated abstractions instead of shipping features.

MCP (Model Context Protocol) solves this by giving AI tools direct access to your Courier account. Cursor can see your templates, check delivery logs, and generate code that matches your actual configuration.


Step 1: Connect Courier MCP to Cursor

Add the MCP server to your Cursor config:

Copied!

// .cursor/mcp.json
{
"mcpServers": {
"courier": {
"url": "https://mcp.courier.com",
"headers": {
"api_key": "YOUR_COURIER_API_KEY"
}
}
}
}

For Claude Code:

Copied!

claude mcp add --transport http courier https://mcp.courier.com --header api_key:YOUR_COURIER_API_KEY

Get your API key at app.courier.com/settings/api-keys.

What changes once MCP is connected: Cursor can now query your actual Courier account. It sees your templates, users, and delivery history. When you ask it to "send a password reset notification," it checks which templates exist and generates code using the right template ID—not a placeholder.

Full MCP setup guide →

Claude Code with Courier MCP


Step 2: Validate Your Setup with the CLI

Before asking Cursor to generate implementation code, verify that your Courier configuration works. The CLI lets you test notifications without writing code.

Copied!

npm install -g @trycourier/cli
courier login

Send a test notification:

Copied!

courier send \
--user user_123 \
--title "Test notification" \
--body "Testing from CLI"

Check delivery logs:

Copied!

courier logs --user user_123

List your templates:

Copied!

courier templates list

If notifications deliver successfully from the CLI, your provider configuration is correct. Now Cursor can generate code with confidence that the underlying setup works.

CLI reference →


Step 3: Prompt with Context

With MCP connected and your setup validated, Cursor has everything it needs. Here's how to prompt effectively:

Cursor with Courier MCP

Check what you have: > "Show me my notification templates in Courier. Which one handles order confirmations?"

Build against existing templates: > "Create a function that sends the ORDER_SHIPPED template to a user. Include the tracking URL in the data payload."

Debug delivery issues: > "The notification to user_456 didn't arrive. Check Courier logs and tell me what failed."

Add new templates: > "Create a Courier template called COMMENT_NOTIFICATION for when someone comments on a post. Send to push first, fall back to email."

The key difference from prompting without MCP: Cursor validates against your actual account. If you reference a template that doesn't exist, it tells you. If you ask for a channel you haven't configured, it flags it.


What You Can Build

Once MCP is connected, here's what becomes straightforward to implement:

In-App Notification Inbox

A notification center with unread badges, real-time updates, and mark-as-read functionality. Courier provides drop-in React components that handle the WebSocket connections and state management.

React SDK documentation →

Mobile Push Notifications

iOS and Android push with automatic token management, badge counts, and deep linking on tap. The mobile SDKs handle provider registration (APNs, FCM) and sync tokens to Courier automatically.

iOS SDK documentation → Android SDK documentation → Flutter SDK documentation → React Native SDK documentation →

Slack and Teams Integration

Send notifications to users' Slack DMs or Teams channels. Courier handles the OAuth flow, user mapping, and Block Kit formatting for rich messages.

Slack channel notifications

Slack integration guide → Teams integration guide →

User Preference Management

Let users control which notifications they receive and on which channels. Courier provides a drop-in preference center component and API for custom implementations.

Preference management UI

Preferences documentation →

Scheduled and Automated Notifications

Reminders, drip sequences, and event-triggered workflows. Set up automations that send a series of notifications based on timing or user actions.


When This Approach Works Well

The MCP + CLI + Cursor workflow is most effective when:

  • You're adding notifications to an existing app (not building notification infrastructure as the product)
  • You want multi-channel delivery without managing multiple provider SDKs
  • Your team includes non-engineers who need to edit templates
  • You're iterating quickly and want to test changes without redeploying
  • You need sophisticated notification system, but don't want to use up engineering time

It's less suited for:

  • Cases where you need to self-host everything for compliance reasons
  • Simple single-channel use cases where a direct provider SDK is sufficient

Comparing Notification Platforms for AI-Assisted Development

If you're evaluating options, here's how the main platforms compare for AI coding workflows:

Notification platform comparison

Courier

API-first platform that brings together developers, PMs, and marketers. Developers get clean APIs and SDKs. PMs get visual workflow builders. Marketers get template management without touching code.

  • Official MCP server with embedded SDK installation guides
  • Native Slack and Teams support with rich formatting
  • Drop-in components for inbox, toasts, and preferences
  • 50+ provider integrations with automatic failover
  • Free tier: 10,000 notifications/month

Knock

Focuses on the workflow engine layer. Good for teams that want to define notification logic as code.

  • Has an MCP server for workflow/API operations
  • Requires third-party providers for actual delivery
  • Limited Slack/Teams support compared to Courier

Novu

Open-source option. Self-host if you want full control over infrastructure.

  • Has an MCP server for API operations
  • Self-hosting means managing infrastructure yourself
  • Community-driven integrations vary in quality

OneSignal

Comes from mobile push notifications. Strong on push and email for consumer engagement.

  • Community-built MCP server (not official)
  • Campaign-centric model fits marketing better than product notifications
  • Limited B2B channels (no native Slack/Teams)

Resources

ResourceLink
MCP Server Setupcourier.com/docs/tools/mcp
CLI Referencecourier.com/docs/tools/cli
API Documentationcourier.com/docs/reference
SDK Overviewcourier.com/docs/sdk-libraries/sdks-overview
Free Tiercourier.com/pricing

Getting Started

  1. Create a Courier account (free tier includes 10k notifications/month)
  2. Configure your first provider (email is easiest to start)
  3. Connect MCP to Cursor
  4. Start prompting

Related guides:

Similar resources

b2b customer engagement guide
Notifications LandscapeCourier

The Complete Guide to B2B Customer Engagement

Courier provides the notification infrastructure layer for B2B customer engagement, routing messages across email, SMS, push, in-app, Slack, and Teams based on user preferences and product events. Unlike building notification systems in-house—which takes months of engineering time for features like multi-channel routing, preference management, and delivery tracking—Courier handles this infrastructure so product teams can focus on engagement strategy. B2B customer engagement requires multiple layers: notification infrastructure (Courier), customer data platforms (Segment), product analytics (Mixpanel/Amplitude), and channel-specific tools. Companies with strong engagement programs see 15-25% churn reduction. The key is connecting product events to customer communication at the right moment through the right channel, handling complexity like multiple users per account with different notification needs across work channels.

By Kyle Seyler

January 20, 2026

what is a customer engagement platform
Product ManagementNotifications LandscapeCourier

Customer Engagement Platform vs CRM: Key Differences Explained

A CRM stores customer data: contacts, purchases, support tickets, and pipeline. It answers "who are our customers?" A customer engagement platform (CEP) orchestrates communication across email, push, SMS, in-app, and chat. It answers "what should we tell them next?" CRMs focus on historical records. CEPs process real-time behavior and trigger messages based on actions. Most teams need both, plus a third layer: notification infrastructure for reliable multi-channel delivery. Courier bridges CEP and infrastructure by combining routing, failover, and delivery tracking with engagement features like preference management, visual templates, and in-app notification centers.

By Kyle Seyler

January 07, 2026

investigation workflow
CourierEngineering

How We Investigate Support Tickets at Courier

Courier's support team resolves complex issues 4X faster using parallel investigation. Here's how it works: when a ticket comes in, an AI agent starts exploring the codebase while the support engineer examines actual customer data. The agent traces code paths and searches past investigations. The human reads event logs and forms hypotheses based on real state. Running both simultaneously catches mismatches fast—the agent sees what could cause a problem, the human sees what actually happened. This post breaks down the workflow, tools, and documentation structure that makes it repeatable.

By Thomas Schiavone

December 18, 2025

Multichannel Notifications Platform for SaaS

Products

Platform

Integrations

Customers

Blog

API Status

Subprocessors


© 2026 Courier. All rights reserved.