LoopX is a provider-neutral control plane for long-running AI agent work, designed to keep goals, evidence, quotas, and handoffs stable across many bounded turns.
LoopX is a lightweight, stateful layer for managing long-running agent workflows without replacing the runtime that actually performs the work. The README presents it as open and provider-neutral, meant to coordinate agents such as Codex, Claude Code, Cursor, or a custom runtime while preserving durable objectives, gates, todos, evidence, quota, and handoffs. It is positioned for loop engineering across engineering, research, operations, and team-based agent work.
The project addresses the difficulty of governing work that lasts beyond a single session. According to the README, chat memory and a timer are not enough when objectives change, owner decisions appear, evidence becomes stale, work must pass between agents, or execution should stop once there is no useful next step. LoopX aims to make long-running work reviewable, restartable, and easier to hand off while keeping human judgment in the loop for risky or final decisions.
Conceptually, LoopX keeps the durable control state in one central layer and lets an agent runtime execute only a bounded slice of work at a time. The README describes a loop where LoopX checks whether human judgment is needed, whether a safe fallback exists, and whether quota allows another tick; then a chosen agent performs one turn, after which evidence, handoff information, and the next todo are written back into state. It also frames the system as an agent-native Kanban: the board is only a view, while LoopX state remains the source of truth, and registered agents act as peers with claims, leases, task boundaries, capabilities, and typed continuation deciding who goes next.
It is gaining attention because the repository sits at the intersection of long-running AI agents, workflow automation, and practical agent operations. The README emphasizes provider-neutral coordination across popular coding agents, durable state for multi-day work, and public evidence arcs that show real loops rather than one-turn demos. The star growth in the provided metadata also suggests that the project is resonating right now, especially with people exploring agent control planes, handoffs, and governed automation.
The README does not name direct competitors, but it clearly contrasts itself with chat memory, timers, and provider-specific orchestration runtimes. By positioning LoopX as not another agent framework, it appears closer to a control-plane or state-kernel approach than to a full agent framework. Compared with generic workflow automation, its distinguishing focus is on durable agent governance: goals, evidence, quota, gates, and handoff state for long-running AI work.
AI-explained · grounded in each repo's README