Loop engineering state kernel
The agent restarts. The goal does not.
loopx is a lightweight loop engineering state kernel for long-running AI agent teams. It holds durable goals, executable todos, evidence logs, and verifiable handoffs outside any single agent process, so the loop survives a context reset.
Agent-loop agnostic: CodexClaude Codeother coding agents
01
Durable goals
The objective lives in the kernel, so losing a session does not lose what the team was trying to do.
02
Quota-aware auto-wake
The loop knows when it ran out of budget and when it can pick the work back up.
03
Evidence logs
Each step leaves a record, which is what makes the next agent able to trust the last one.
04
Verifiable handoffs
Work moves between agents with something checkable attached instead of a summary.