Pushary

Approve your AI agents from your phone in one tap

AILaunched Jul 2026

The story

Pushary is a control panel for AI agents that lets you approve or deny permission requests from your phone, Slack, or web app. Instead of agents freezing while waiting for your decision, Pushary delivers requests to you wherever you are and lets you respond in seconds. Your code never leaves your machine - only the decision itself is sent to Pushary's servers.

Overview

Pushary is a control panel designed for developers running AI agents like Claude Code, Cursor, Codex, and any MCP-compatible agent. It solves the problem of agents freezing when they encounter steps requiring human approval, such as deploying code, deleting files, or making other irreversible changes. Instead of waiting hours for you to return to your machine, Pushary sends permission requests to your phone, Slack, or web app, allowing you to approve or deny them in seconds. You can set permission policies once to auto-approve safe, reversible actions while escalating risky ones (deletions, deployments, spending) to your phone. The system maintains a full audit trail of every decision and integrates with n8n workflows and Slack for multi-channel approvals. Setup takes under two minutes with a single command, and your source code never touches Pushary's servers - only the decision request itself is transmitted.

Key features

Phone notifications

Receive permission requests on your phone and approve or deny them from your lock screen in one tap.

Permission policies

Set rules once to auto-approve safe, reversible actions while escalating risky ones like deletions and deployments.

Multi-channel delivery

Decisions reach you via phone, Slack, web app, and audit trail so you stay informed across all platforms.

Works with any MCP agent

Integrates with Claude Code, Cursor, Codex, Hermes, Windsurf, and any agent using the Model Context Protocol.

Audit trail

Every permission request and decision is recorded and synced for full visibility and compliance.

n8n integration

Add human checkpoints to any n8n workflow to pause, ask for approval, and continue with your answer.

Use cases

1
Long-running development tasks
Hand off refactoring or testing tasks to Claude Code and approve changes from your phone while away.
2
Automated deployments
Let agents handle routine deployments but require your approval for production changes before they go live.
3
Workflow automation
Add human approval gates to n8n workflows so critical steps wait for your decision instead of auto-executing.
4
Multi-agent management
Monitor and control 6+ agents from one dashboard, seeing which ones need your attention and their full history.

FAQ

Can I approve an AI agent from my phone?

Yes. Pushary sends permission requests to your phone and you can approve, deny, or choose options from your lock screen in one tap.

Is my code sent to Pushary's servers?

No. Your source code never leaves your machine. Pushary only receives the permission request itself - the title, body, and tool name.

How long does setup take?

Setup takes under two minutes. Run one command (npx @pushary/agent-hooks setup) and Pushary configures your MCP server, permission hooks, and API key.

What AI agents work with Pushary?

Pushary works with Claude Code, Cursor, Codex, Gemini CLI, Hermes, Windsurf, and any agent using the Model Context Protocol (MCP).

What happens if I don't answer a permission request?

Your permission policy answers for you. You can set each rule to auto-deny or auto-approve after a timeout (default 10 minutes), or hand control back to the terminal.