PubNub MCP Server

PubNub MCP Server

A hosted Model Context Protocol (MCP) server that exposes PubNub SDK documentation and PubNub API resources to LLM-powered tools. This improves the LLM AI Agent's ability to understand and interact with PubNub's SDKs and APIs. Uses HTTP transport with OAuth authentication.

Features

• 📚 Comprehensive SDK Documentation - Access detailed documentation, code examples, and implementation guides for 20+ programming languages including JavaScript, Python, Java, Swift, Kotlin, C#, Ruby, Go, and more
• 🏗️ Application & Keyset Management - Create, configure, and manage PubNub applications and keysets with features like message persistence, file sharing, presence tracking, and app context
• 💬 Real-time Communication - Send and receive messages across channels, implement live chat, notifications, and real-time updates with support for both messages and lightweight signals
• 👥 User & Channel Management - Manage user profiles, channel metadata, and membership relationships with full CRUD operations for building community and social features. In addition, there is an option to view & manage user profiles, channels and memberships in PubNub Admin Portal (this requires Manage Premium).
• 📍 Presence & Activity Tracking - Monitor real-time user presence, see who's online in channels, and track user activity across your application
• 📊 Real-time Decisioning & Analytics - Trigger the right action when it matters, measure its impact, and adjust as conditions change -- all in real-time. Requires Illuminate.
• 📈 Channel, User & Message Analytics - Spot patterns in aggregated app activity, from top channels and country breakdowns to new vs. recurring users and device distribution, to understand engagement, usage, and performance. In addition, limited metrics are available for viewing in PubNub Admin Portal without Manage Premium; full analytics require Manage Premium.
• 🔍 Usage & Monitoring — Track billable usage across apps and keysets, review current and historical consumption, and export reports to understand cost drivers.
• 🔧 Multi-Platform Integration - Works with Cursor, Visual Studio Code, Claude Code, and other MCP-compatible AI assistants
• ⚡ Developer Experience - Built with TypeScript for type safety, includes testing infrastructure

Quick Start

The hosted server is managed by PubNub and provides the easiest way to get started — no installation required, and authentication is handled automatically via OAuth. If your AI tool doesn't support remote MCP servers, use the locally installed version (npx @pubnub/mcp@latest) instead.

Note: To switch to the hosted server (https://mcp.pubnub.com), remove the old local configuration from your AI client first, then follow the setup steps below.

Hosted PubNub MCP Server

Connect your AI assistant to https://mcp.pubnub.com — no installation required. Authentication is handled automatically via OAuth.

VS Code

Click the button above, then select Open in Visual Studio Code. Back in VS Code, click Install, then select the organization you want to authorize and click Allow access.

Alternatively, add the following to your VS Code settings.json manually:

{

"mcp": { "servers": { "pubnub": { "url": "https://mcp.pubnub.com" } } } }

Learn more in VS Code documentation

Cursor

Click the button above, then select Open Cursor. In Cursor, click Install, select the organization you want to authorize, and click Allow access. Navigate to Cursor Settings → Tools & MCP to verify PubNub is enabled.

Alternatively, add the following to .cursor/mcp.json (or ~/.cursor/mcp.json for global configuration) manually:

{

"mcpServers": { "pubnub": { "url": "https://mcp.pubnub.com" } } }

When you save the file, a notification is displayed. Click Enable to activate the MCP server. Learn more in Cursor documentation

Claude Code

In the terminal, run:

claude mcp add --scope user --transport http pubnub https://mcp.pubnub.com

Then run claude to open Claude Code and enter /mcp. Select pubnub, then authenticate, and click Allow access to complete authorization. Learn more in Claude Code documentation

Claude Desktop

Note: The following instructions apply to Pro and Max plans. For Enterprise plans, the setup is configured by your organization administrator. Refer to Anthropic's documentation for details.

1. In Claude Desktop, navigate to Customize → Connectors.
2. Click + and select Add custom connector.
3. Enter https://mcp.pubnub.com as the connector URL.
4. Complete the OAuth login when prompted.

Learn more in Claude Desktop documentation

Codex

In the terminal, run:

codex mcp add pubnub --url https://mcp.pubnub.com

Then run codex mcp login pubnub to authenticate, select the organization you want to authorize, and click Allow access. Learn more in Codex documentation

Codex Desktop

1. Navigate to Settings → MCP servers and click + Add server.
2. Select Streamable HTTP as the transport type.
3. Enter https://mcp.pubnub.com as the server URL.
4. Follow the OAuth login prompt to complete authorization.

Learn more in Codex Desktop documentation

Gemini CLI

In the terminal, run:

gemini mcp add pubnub --scope user --transport http https://mcp.pubnub.com

Then run gemini to open Gemini CLI and run /mcp auth pubnub to authenticate. Learn more in Gemini CLI documentation

OpenCode

Add the following to ~/.config/opencode/config.json:

{

"mcp": { "pubnub": { "type": "remote", "url": "https://mcp.pubnub.com", "enabled": true } } }

Then run opencode mcp auth pubnub to authenticate. Learn more in OpenCode documentation

Local PubNub MCP Server

If your AI tool doesn't support remote MCP servers, run the server locally with:

npx @pubnub/mcp@latest

API Key

Before you begin, we highly recommend creating a Service Integration in the PubNub Admin Portal and providing your API key to the MCP server. While some basic features will work without it, adding an API key unlocks much more functionality. Alternatively, refer to Local server configuration for instructions on configuring the server to work with a single PubNub keyset.

The installation process for an MCP server depends on the AI assistant you're using. For the standard setup, you'll need Node.js (v20.0.0 or higher).

VS Code

Just click the link above, then select "Open in Visual Studio Code" on the page that appears. Back in VS Code, click "Install". You'll be prompted to enter your PubNub API Key. Once provided, your MCP server is ready to use. For additional configuration options, refer to Local server configuration.

Cursor

Click the link above, then select "Open Cursor" on the page that appears. Back in Cursor, there's a "Install MCP Server?" prompt. Make sure to provide the value for variable holding your PubNub API Key. Once you do, click "Install". Your MCP server is now ready to use. For additional configuration options, refer to Local server configuration.

Claude Code

With Claude Code installed run this command to have the MCP added to your configuration. Make sure to replace the value of :

claude mcp add pubnub --env PUBNUB_API_KEY=<your-api-key> --scope user --transport stdio -- npx -y @pubnub/mcp@latest

Server is added in the "User" scope which means it will be available across all projects. For additional configuration options, refer to Local server configuration.

Codex

With Codex installed run this command to have the MCP added to your configuration. Make sure to replace the value of :

codex mcp add pubnub --env PUBNUB_API_KEY=<your-api-key> -- npx -y @pubnub/mcp@latest

For additional configuration options, refer to Local server configuration.

Gemini CLI

Gemini CLI does not support automatic MCP installations. You'll have to manually edit your settings.json file and add the section below. Make sure to replace the value of :

"mcpServers": {

"pubnub": { "command": "npx", "args": [ "-y", "@pubnub/mcp@latest" ], "env": { "PUBNUB_API_KEY": " " } } }

For additional configuration options, refer to the Local server configuration section in the PubNub documentation.

Development

Prerequisites

• Node.js 22.23.4, 24.22.1, or 26.12.1, or a later release on those lines

Running locally

npm install

npm run dev

Environment variables

Testing

npm run test:unit