Novu

Novu is an open-source web app and server that provides communication infrastructure for products and AI agents. It acts as a unified layer to manage notifications across multiple channels, including email, SMS, push notifications, and chat, using a single API. The software is deployed as a server and web application to centralize how applications talk to their users.

The platform serves two primary functions. For product teams, it manages the delivery of transactional and marketing notifications through a centralized dashboard and API. For developers building AI agents, it provides an Agent Communication Infrastructure that normalizes inbound and outbound messages across different platforms. This removes the need to maintain separate webhook handlers for every single integration, allowing agents to interact with users on the channels they already use.

Key features

• Unified API for multiple messaging providers across email, SMS, push, and chat
• Embeddable real-time notification center inbox component
• Workflow engine with support for branching and conditional logic
• Digest engine to batch multiple notifications into a single message
• No-code email editor for template creation
• Embeddable preferences component for user-controlled notification settings
• Bidirectional messaging for AI agents with a unified conversation model
• Integrations with Slack, Microsoft Teams, Telegram, and WhatsApp

Novu is built with TypeScript, React, and Node.js. It supports a wide array of providers, including 19 email, 37 SMS, 8 push, and 12 chat services. The architecture allows developers to bring their own agent logic, whether built with LangGraph, AI SDK, or custom stacks, while Novu handles the delivery and normalization of messages. This ensures that messages from different platforms are converted into one consistent shape before reaching the agent and formatted correctly when sent back to the user. It is designed for developers who need to scale communication without rebuilding provider integrations from scratch.

The project follows an open-core model, where the core technology is licensed under the MIT license while specific enterprise features require a commercial license.