Agent Approve, an Oakland-based startup founded in January 2026, publicly launched its iOS and Apple Watch app today. The app provides a single mobile interface for developers to monitor and approve commands across 14 coding agents, including Claude Code, OpenAI Codex, Cursor, OpenClaw, GitHub Copilot, Hermes, Google Antigravity, OpenHands, Windsurf, OpenCode, and Pi, according to the company’s PR Newswire announcement.
The Problem: Terminal Babysitting vs. YOLO Mode
The product targets a specific tension in agent-assisted development. Agents running autonomously can execute destructive commands (wiping production databases, force-pushing over weeks of work) when left unchecked. But requiring terminal presence for every approval creates idle time that defeats the purpose of autonomous agents.
“Step away from the terminal, and an agent sits idle waiting for an approval. Turn on yolo mode, and you risk wiping your home directory,” founder Jim Beno said in the announcement.
How It Works
Setup runs through npx agentapprove, which installs hooks into supported agent platforms. Developers pair the app by scanning a QR code. When an agent requests approval, the full command and its context appear as a push notification on iPhone or Apple Watch. One tap approves or denies, and the agent resumes.
The app includes a “Restrictive” policy mode that auto-blocks over 250 destructive command patterns. Compound commands are parsed individually, so a dangerous operation cannot hide inside a chain of safe ones. Developers can also configure custom auto-approve and deny rules, building policy incrementally as they work. Voice commands let developers send follow-up instructions to agents through the app’s microphone, according to the product website.
Pricing and Privacy
Agent Approve costs $14.99 per month with a 7-day free trial. The company offers configurable privacy tiers with selectable data retention periods and an end-to-end encryption option where developers control their own keys.
A Growing Control Layer Category
Agent Approve enters a week that has already seen significant movement in the agent governance space. Microsoft launched Execution Containers (MXC) for policy-driven agent containment on Windows. Radware expanded its Agentic AI Protection with Claude Code endpoint monitoring. The SkillCloak research showed that malicious agent skills evade major scanners at over 90% rates.
Agent Approve’s bet is narrower: rather than enterprise-grade governance frameworks, it targets individual developers and small teams who want mobile oversight without changing their existing workflows. Whether a $15/month app can compete with platform-native safety features that agent vendors are building themselves remains the open question.