Kule needs testers: run AI agents on a copy of your folder, then tell me what broke

Description and use

Kule is a sandboxed control room for AI agents working on folders. The author requests users to run it locally, report unexpected behavior, and help refine policies and isolation.

Problem

Kule is a new open‑source control room for AI agents that work on folders, but the author needs real‑world testing to validate its safety and usability.

Outcome

The author invites users to run Kule on their own machines, report unexpected behavior, and help refine policies and sandboxing. Early feedback will drive fixes in subsequent previews.

Workflow

• 1. Install Kule via the install guide.
• 2. Create a test folder or use a committed project.
• 3. Run one of the six testing scenarios (install check, loop without agent, app with agent, own work, own setup, boundaries).
• 4. Observe the sandboxed actions, policy prompts, and terminal output.
• 5. Report findings by opening a GitHub issue (Bug report, Test report, or Idea).

Evidence

Kule is a control room for AI agents that work on your folders. You point an agent (Claude Code, Codex, Gemini, or any agent that speaks ACP) at a folder.

Source

The default policy is an allow‑list now.

Source

Nothing an agent writes reaches your home.

Source

Your terminals run in the task's sandbox too, not just the agent.

Source

I wrote a short testing guide with six scenarios.

Source