#299 · Primary category: AI Agents & Automation

loopx

agent-control-plane agent-harness agent-ops ai-agents codex long-horizon long-running-agents loop-engineering loopx workflow-automation

Long-horizon agent control plane for durable, governed work across Codex, Claude Code, and other harnesses.

Project last updated:08/30/26

GitHub Stars

5.3K

Forks

476

Contributors

59

License

Apache-2.0

Why we included this project

Agent work that spans days rather than a single session has a way of losing the thread between turns: objectives drift, scope blurs, and evidence gets buried in chat history. LoopX is a local-first state kernel that sits alongside runtimes like Codex, Claude Code, or Cursor and keeps the durable control state (objectives, gates, todos, evidence, quota) in one compact layer, so long-running engineering, research, and operations work stays reviewable and restartable. It is deliberately not another agent framework or a provider-specific orchestrator; the runtimes still execute the work, while LoopX governs who acts next, when a human decision is needed, and when a bounded agent slice is a safe fallback. Teams running multi-day issue-fix arcs or experiment loops will get the most from the durable goals and quota-aware auto-wake, and the verifiable handoffs make it safe to pass work between agents. There's also a developer book and reproducible walkthroughs, so you can test the loop against your own workflow before adopting it.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category