Agent Q is a research framework developed by MultiOn (now rebranding as AGI, Inc.) that combines guided Monte Carlo Tree Search (MCTS), AI self-critique, and Direct Preference Optimization (DPO) to enable autonomous agents to plan, execute, and self-correct on complex web tasks. It demonstrated a 440% improvement in zero-shot booking task performance. By 2026, MultiOn has shifted focus toward a mobile AGI-0 app while maintaining the developer API for web automation. Key features: - Monte Carlo Tree Search for multi-step web task planning - Self-supervised improvement via DPO and AI self-critique - Real-browser execution: clicks, form fills, navigation, scrolling - REST API for integrating web-action capability into other applications - Asynchronous and synchronous task execution modes
Not public. Trial tier available for new users. Transparent pricing not currently listed on the multion.ai site.