OneSignal MCP Server

The OneSignal MCP Server is an official implementation of the Model Context Protocol (MCP) that enables developers and marketing teams to seamlessly integrate OneSignal features into MCP-capable AI clients such as Claude Desktop, Cursor, ChatGPT, or GitHub Copilot. This standardized integration eliminates the need to constantly switch between code editors, terminals, and the OneSignal web dashboard. Users can launch notification campaigns, manage audience segments, and perform delivery analyses directly through natural language. The official server is available both as a managed remote service via secure OAuth authentication and locally through a Smithery CLI installation package.

What is the OneSignal MCP Server

The Model Context Protocol decouples AI assistants from the need to manually integrate proprietary APIs. Without the OneSignal MCP Server, a developer would have to write custom code or copy API payloads into the prompt for every task – whether triggering a push notification or querying click-through rates. Instead, the MCP server provides the AI with a structured directory of predefined tools. The AI independently detects when and how to call each tool, validates input parameters automatically, and returns the result to the conversation context. This significantly simplifies both development and marketing workflows.

Features: Messaging and Audience Management via AI

According to the provider, the server exposes a variety of tools that directly access the OneSignal REST API. The core capabilities include:
Message Delivery: Push notifications, emails, and SMS can be sent directly from the chat to predefined segments or specific test devices. The AI can load templates and insert dynamic personal variables.
Subscriber Management: User profiles and their subscription statuses can be searched, viewed, and updated. New identities and device-specific aliases can be assigned directly.
Audience Segmentation: Target groups can be created and named based on custom filters (such as inactivity or app version), for example: "Create a segment for users who have not opened the app for three days".
Campaign Analytics: Delivery statistics, open rates, and click-through data of current campaigns can be queried and compared using simple conversational commands.

Setup and Configuration

For most AI clients, OneSignal recommends using the official hosted remote server. The connection is established via a secured authentication interface. In Cursor or Claude Desktop, the server is registered using the official connection URL https://api.onesignal.com/mcp/oauth. Upon initiating the connection, a secure OneSignal login page opens, allowing the user to sign in and grant the AI client access to the selected applications. For developers who prefer a local configuration, a Smithery installation path is also available via npx -y @smithery/cli install onesignal-mcp --client claude.

Security and Privacy according to the Provider

Since the OneSignal MCP Server accesses sensitive user data and communication channels, the provider recommends strict security guardrails. Authentication via OAuth ensures that the AI client never needs to store raw API keys or master credentials in plain text. Additionally, the official implementation features built-in confirmation barriers (Confirmation Guards). For wide-reaching mutations – such as broadcasting a message to a large audience – the system requests explicit human confirmation. This prevents the AI from accidentally sending mass notifications.

Benefits and Evaluation

The OneSignal MCP Server provides a massive productivity boost for teams managing mobile and web applications. It translates complex API integration steps into simple conversational patterns. While community developers maintain additional unofficial Python implementations on GitHub (such as weirdbrains/onesignal-mcp) offering up to 57 tools, the official server from OneSignal stands out due to its native OAuth support and direct maintenance by the vendor, ensuring long-term compatibility with future API updates.