Planarian: Managing Agent State with Statepoints

2026-09-28

AI Summary

This study addresses the absence of a unified and safe recovery mechanism for LLM agents modifying local and remote states by proposing Planarian, a novel runtime system. Planarian introduces the abstraction of state points, pioneering an incremental file system snapshotting and remote compensating action logging mechanism that operates without external service dependencies. This enables unified management, rollback, and branching exploration across heterogeneous environment states. By leveraging transparent compensation logging and parallel branch isolation techniques, Planarian improves task quality by up to 15× while incurring only a 3% overhead for users recovering from erroneous operations.

Abstract

LLM agents solve complex tasks by iteratively changing files, invoking local tools, and interacting with remote services, which modifies state across their local environment and remote services. Today, agents and users must manage these changes explicitly, whether reverting exploratory actions or recovering from erroneous ones. Doing so safely requires coordinated actions, yet current agent harnesses lack unified abstractions and mechanisms for managing local and remote state consistently and efficiently. We describe Planarian, an agent runtime with state management that enables agents and users to recover from erroneous actions and explore alternative executions over consistent local and remote environment state. Planarian introduces the abstraction of agent statepoints, which are consistent, restorable point-in-time versions of the environment state. Planarian exposes three state-management primitives to agents and users: (i) snapshot creates a new statepoint spanning local and remote state without requiring external services to support checkpoints: it relies on efficient incremental process and file system snapshotting to capture local sandboxed state, and transparently records compensating actions to undo remote state changes; (ii) rollback restores the environment to a previous statepoint by reverting to a prior local checkpoint and replaying compensating actions for remote state changes; and (iii) fork creates multiple isolated branches from a statepoint, enabling the agent to explore alternatives in parallel. We show that Planarian enables agents to undo mistakes and explore alternatives in parallel, improving task quality by up to 15x, and allows users to recover from erroneous actions with only 3% overhead.

Problem

• LLM agents
• state management
• local and remote state
• error recovery

Innovation

• LLM agents
• state management
• statepoints
• compensating actions
• agent runtime