# Agent Push Kit [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/JustinBStrong/agent-push-kit-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 138  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/agent-push-kit

## Description
Send, search, and manage notifications, accounts, and push preferences

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "agent-push-kit": {
    "command": "npx",
    "args": ["-y","agent-push-kit-mcp"],
    "env": {
      "AGENT_PUSH_KIT_TOKEN": ""
    }
  }
}
```

**Requires environment variables:** `AGENT_PUSH_KIT_TOKEN` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What Agent Push Kit MCP server does

The Agent Push Kit MCP server exposes Agent Push Kit functionality through the Model Context Protocol. It is intended for MCP-compatible AI clients that need to work with notification data and account-level settings without using a separate custom integration.

Supported operations cover three areas:

- Sending notifications
- Searching notifications
- Managing accounts and push preferences

The available material identifies the service and operation groups, but does not provide individual tool names, argument schemas, return formats, or examples for each operation.

## How it works

Agent Push Kit MCP server runs as an MCP process using stdio. An MCP client starts the process and communicates with it through the standard input and output streams. Authentication is supplied through the `AGENT_PUSH_KIT_TOKEN` environment variable.

The token should be a named token created through Agent Push Kit’s agent connection page. Once the process is launched with that token, the client can access the server’s supported notification, account, and push-preference operations.

## Setup and configuration

Create a named token at [agentpushkit.com/connect-agent](https://agentpushkit.com/connect-agent), then provide it when launching the server:

```bash
AGENT_PUSH_KIT_TOKEN=apt_YOUR_TOKEN npx -y agent-push-kit-mcp
```

The package is started with `npx`, so the command can retrieve and run the MCP package without requiring a separate installation step in the documented example. Add the command and environment variable to the MCP client’s stdio configuration according to that client’s setup format.

For implementation details and service-specific configuration guidance, consult the [Agent Push Kit MCP documentation](https://agentpushkit.com/docs/mcp).

## Tools and capabilities

Agent Push Kit MCP server is relevant when an agent needs to perform notification workflows through Agent Push Kit. The documented capability groups are:

- Send notifications to the connected Agent Push Kit service.
- Search existing notifications.
- Manage Agent Push Kit accounts.
- Read or change push preferences.

The supplied README does not specify whether search supports filters, how accounts are identified, which notification fields are required, or what preference settings are available. Those details should be verified in the linked documentation before building an agent workflow around them.

## Limitations and notes

The provided material describes MCP stdio access but does not list supported client-specific configuration files, individual tool schemas, response formats, rate limits, or permission scopes. It also does not state a software license. Treat the named token as a required credential and avoid exposing it in client logs or shared configuration.

Agent Push Kit MCP server is a fit for clients that can launch standard stdio MCP processes and provide environment variables. The documented launch command uses a token beginning with `apt_`; obtain the actual value from Agent Push Kit rather than using the placeholder shown in the example.

_Full upstream README: https://allmcps.com/mcp/agent-push-kit/readme_

