LoopX
A local-first control plane for long-running agents: goals, decisions and status persist across Codex and Claude Code sessions
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- The tool manages the agent's long-running background tasks and stores their state locally
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/loopxDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Install the tool
pipx install loopxWithout pipx, pip install loopx works too.
Install with python3 -m pip install --upgrade loopx, then run loopx workflow-skills --install and loopx doctor to verify the setup. Restart the agent so it picks up the new workflow skills.
Other ways from the author
python3 -m pip install --upgrade loopx
loopx workflow-skills --install
loopx doctormacOS and Linux, requires an active Python 3.11+ environment.
This is third-party code. Review the repository files before installing.
What it does
LoopX does not replace the agent itself; it runs on top of an already installed runtime, such as Codex or Claude Code, and holds the goal, decision history and confirmed results for a project. After installing with loopx workflow-skills --install, the tool adds workflow skills that let the agent continue a multi-hour task after a restart without losing the previous step's context. A local Personal Agent Workspace dashboard shows what is running now, what needs the owner's decision, and what is scheduled, plus update and rollback commands.
Who it is for. For anyone giving a coding agent long, multi-step tasks who wants progress and decisions to persist across session restarts.
Good fit when
- The agent's task spans many hours or sessions and losing progress is not an option
- You want one place to see what the agent has done and what needs your decision
- Several agents on different runtimes work on the same project in turn
Not a fit when
- The task is short and fits in a single agent session
- You do not want an extra Python tool running alongside the agent itself
Example request
Continue the task from the previous session and show what is already done and what is leftLimitations
Requires Python 3.11 or newer and Node.js 22.18 or newer, since the TypeScript core starts automatically alongside the CLI. The agent must be restarted after install to pick up the new workflow skills. LoopX itself provides no model; that comes from the agent's already configured runtime.
How to disable. Remove the package with pip uninstall loopx; the workflow skills in the agent's folder can be removed by hand if no longer needed.
Security check
- The tool manages the agent's long-running background tasks and stores their state locally
README in short
The README describes LoopX as an open local-first control plane for long-horizon personal agents: goals, decisions and confirmations persist across sessions and restarts. It includes a Personal Agent Workspace with goals, tasks, schedules and recovery, plus reported LHTB benchmark gains over plain Codex. Installation is via PyPI without cloning the repository. Apache 2.0 licensed.
FAQ
Is LoopX a standalone agent?
No, LoopX provides no model of its own and does not replace the agent; it runs on top of an already configured runtime, such as Codex or Claude Code, and tracks the goal and task status.
Which runtimes does LoopX already support?
The README lists Codex, Claude Code and DeepSeek Harness; the list of supported runtimes may grow.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent