Free: $0, includes phone app, MCP server, REST API, common tasks with no model, scheduled tasks, iPhone through Mac (beta). Pro: $5/month founding offer ($8/month regular, $79/year), includes remote access, Claude and ChatGPT apps, cron jobs, recipe sync across up to 3 phones. Power: $29/month, ever
Open Source: free and self-hosted. Free: $0/month hosted plan with up to 5 members. Scale: $20 per user per month. Enterprise: Custom annual contract.
Key features
See and control any app without API
On-device model runs locally on phone
40 built-in tasks with no model
Screen model 1.4 GB for app interactions
Works with Claude, ChatGPT and other models
Remote access through relay
Scheduled tasks and webhooks
iPhone support via Mac app
Payment approval requires user tap
Recipes learned and replayed locally
Tool call interception for all agent frameworks
Automatic approval rules to prevent fatigue
AI judge for complex rule evaluation
Works with LangGraph, CrewAI, Mastra, Claude Agent SDK and more
Open Agent Approval Protocol for framework independence
Self-hosted or cloud-hosted deployment options
Custom permission policies and organization controls
What makes it different
Runs on-device so screen never leaves the phone
NPU acceleration on recent Snapdragon phones
No app installation needed on iPhone
Works with any model via standard MCP
Cloud phones with founding cohort pricing lock-in
Framework-agnostic: works with any AI agent through the open Agent Approval Protocol
No vendor lock-in: open source spec and SDKs let you swap providers without rewriting agents
Comprehensive audit trail for accountability and compliance
Meets reviewers where they work with multi-platform notifications
Policy-based routing to eliminate approval bottlenecks
Opposable
withHuman