Ordalin
Submit a tool

Reviewed tool profile

Pushary

Approve, deny or answer your AI coding agents' permission requests from your phone or Mac, with per-tool rules and an audit trail.

Sources checked

Visit website ↗
Pricing
Paid plans with a 3-day free trial. Agent is $9.99 a month with 5,000 notifications, Agent Pro $19.99 a month with unlimited notifications and budgets, and Team $59 a month for 5 people. Partner plans for companies embedding approvals start at $99 a month, and Enterprise is custom. Official source 1 ↗
Available as
iPhone and Android apps, a Mac notch app, Slack and a web app for answering; setup runs on Mac, Windows and Linux with one npx command. Official source 1 ↗
Primary group
Automation & Agents

Also consider

See all 6 similar tools ↓

What it does

Pushary, from Estonian company RalphNex OÜ, is an approval layer for AI agents. When Claude Code, Codex, Cursor, Hermes or another MCP agent reaches a step that needs permission, Pushary sends the question to your phone, Mac notch, Slack or web app so you can approve, deny or answer while away from the computer. Per-tool rules let safe steps run on their own and send risky ones such as deletes, pushes and deploys to you, and every answer is kept in an audit trail. A Partner plan lets companies add the same approvals for their own product's users.

Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗

Features

  • Approvals on your phone. Agent permission prompts arrive as push notifications you can approve, deny or answer, so a long task does not sit waiting while you are away. Official source 1 ↗
  • One-command setup for many agents. npx pushary@latest setup finds Claude Code, Codex, Cursor, Hermes and other agents on the computer and connects them; other MCP clients can be wired by hand. Official source 1 ↗
  • Per-tool permission policies. Auto-approve low-risk steps such as reads, lint and tests, and escalate destructive ones like rm, git push and deploys. Official source 1 ↗
  • Timeout rules. Each step waits as long as you set; with no answer, your rule either denies or hands the question back to the terminal. Official source 1 ↗
  • Control panel and audit trail. One view of every agent and what is waiting on you, with a saved record of each decision. Official source 1 ↗
  • Team record and routing. On the Team plan, answers land on a shared page, rules can route steps such as pushes to infra to a named lead, and the record exports as CSV or JSON. Official source 1 ↗
  • Approvals inside your product. Partner plans let an agent builder ask its own users before a refund, email or delete, branded with the product's name. Official source 1 ↗

Pricing & free limits

Agent

$9.99per month

One person, every agent, 5,000 notifications a month, phone and Mac apps, permission policies, MCP and API access, 30-day history. Official source 1 ↗

Agent Pro

$19.99per month

Unlimited notifications, budgets, priority support, Policy Autopilot suggestions, 90-day history. Official source 1 ↗

Partner

$99–$799per month

Launch, Growth and Scale tiers for companies embedding approvals for their own users, sized by users and approval requests. Official source 1 ↗

Billing details

All paid plans start with a 3-day free trial and can be cancelled anytime. Official source 1 ↗ · Official source 2 ↗

Partner trials cover 100 users and 100 approval requests for 3 days, and there are no automatic overage charges. Official source 1 ↗

Free access & limits

Free access and its limits were not established by the checked official sources.

Best for & limitations

A developer hands Claude Code or Codex a long refactor, walks away, and approves the one risky step from their phone instead of finding the agent stalled later. Official source 1 ↗

An engineering team keeps one shared record of who approved pushes and production migrations across Cursor, Codex and Claude Code. Official source 1 ↗

A company with its own AI agent asks customers to approve refunds or deletes on their phone before the agent acts. Official source 1 ↗

  • Enforcement depends on the agent. On Claude Code, shell commands and file changes are gated, but other tools only reach your phone when Claude Code itself would ask. Official source 1 ↗
  • Not end-to-end encrypted. Approval requests can include commands, diffs and part of a prompt, and synced transcripts use provider-managed encryption rather than end-to-end encryption. Official source 1 ↗
  • No free plan. Every plan is paid after the 3-day trial. Official source 1 ↗
  • No SSO on self-serve Team. SSO, SCIM and enterprise certifications are not part of the self-serve Team plan. Official source 1 ↗
Pushary product screenshot
Website previewFixed desktop viewport · 1440 × 900

Similar tools

Published tools that share its categories.

FastinoOpen-weight decision models, led by GLiDE, that return probabilities over your own outcomes for agent routing, classification and scoring.
Automation & AgentsPaidChecked Nov 6
Visit
RerunNo-code AI coworkers that run your recurring business tasks on a schedule or trigger, on your own ChatGPT or Claude plan.
Automation & AgentsFree trialChecked Nov 6
Visit
CedricAI teammate in Slack, Telegram or Discord that attends meetings and acts across your tools, with approval before anything sends.
Automation & AgentsFree planChecked Nov 5
Visit