Overview
Push notifications, permission approvals, and human in the loop for AI agents
Your AI agent sends a push notification to your phone when a task finishes, an error happens, or a decision is needed. You approve permissions and answer questions from your phone. No need to sit and watch the terminal.
Why use this
- Step away while agents work. Your phone buzzes when the task is done.
- Approve risky actions remotely. Your agent wants to run
rm -rf ./dist? Tap Yes or No from your phone before it runs. - Answer questions on the go. Pick from options or type a reply, right from your phone.
- Get error alerts instantly. Build failed? You know in seconds, not 40 minutes.
How it works
Your agent -> Pushary -> Your phonePushary runs a hosted MCP server. Your agent connects to it and gets tools to notify you and ask you questions. For agents that support hooks (Claude Code, Codex, Cursor, Hermes), Pushary can also step in before a risky action and require your approval first.
Get started
Setup takes about two minutes. One command connects your agent and your phone.
Set up your agent
Claude Code
Native hooks plus MCP tools. Enforced approvals on shell commands, writes, and edits. The deepest integration.
Codex
Native hooks plus MCP tools. Enforced approvals on shell and file edits.
Cursor
A plugin that enforces approvals on risky shell commands, or plain MCP for notifications only.
Hermes
Native Python plugin with automatic error alerts and opt-in approval gating.
Windsurf and other MCP agents
Connect any MCP client for notifications and questions.
Lovable
Connect Pushary as a personal connector so the Lovable Agent can notify you and ask questions while it builds.
No-code and HTTP
Use a plain REST endpoint from n8n, Zapier, scripts, or any language.