Pushary

What is Pushary?

Pushary is a control panel that bridges AI agents and you. When an agent like Claude Code, Codex, Cursor, Gemini CLI, or Hermes needs permission to run a command, change a file, or choose an option, it pauses and sends that decision straight to your phone. Tap approve, deny, pick an option, or type a reply from your lock screen — no need to rush back to the terminal. Pushary also delivers the same request to Slack, the web dashboard, and a full audit trail. Setup takes one command (npx @pushary/agent-hooks setup) and under two minutes. New in this launch: native iPhone and Android apps with QR code pairing, per-tool policies that auto-approve safe reads, and a complete audit log for every decision.

Who it's for

• Developers running AI coding agents – If you hand off refactoring, testing, or batch changes to Claude Code or Codex, Pushary keeps those tasks moving without you chained to the terminal.
• Engineering teams that need oversight – The audit trail and Slack integration let teams see every permission granted or denied, making it easy to review agent actions later.
• Freelancers and solo devs working asynchronously – Start a long task, walk away, and approve the few risky steps from your phone. Come back to finished work instead of a stalled agent.

Key features

Phone notifications

Every decision your agent needs arrives as a push notification on your phone. Approve, deny, or choose from the lock screen — you don’t need to open an app or type a command. The new native iPhone and Android apps make pairing as simple as scanning a QR code shown in the terminal.

Auto-approve safe steps & escalate risky ones

You set the rules once. Low-risk, reversible operations like reading files or making non-destructive edits can be auto-approved. Dangerous actions — deleting, deploying, spending money, anything you can’t undo — are always escalated to your phone. You also set a timeout: if you don’t answer within that window, the rule you configured (deny or return to terminal) takes effect automatically.

Multi-agent dashboard and audit trail

See every agent you run in one web dashboard. Which one is waiting on you? What did it ask? The audit trail records every permission request and your response, across Claude Code, Codex, Cursor, Hermes, Windsurf, and any MCP-compatible agent. No more wondering what got approved while you were away.

What stands out

Your code never leaves your machine. Pushary sends decisions