I built Kule: AI agents work on a copy of your folder, and you decide what comes back

Description and use

Kule lets AI agents work on a copy of a folder, reviewing and approving changes before they affect the real filesystem, providing sandboxing and policy‑based control.

Problem

Coding agents run commands as the user, exposing sensitive files, keys, and network access; users cannot see the actual effects of commands before approval, leading to accidental or malicious changes.

Outcome

Kule enables safe, reviewable AI agent interactions with local folders, preventing accidental or malicious changes by isolating agent actions and requiring policy‑based approval before committing.

Workflow

• Create a task with kule new <folder>
• Run an agent or command with kule exec or kule run
• Review the list of changes and policy verdicts with kule status or kule steps
• Commit approved changes with kule commit or undo with kule undo

Evidence

Kule gives it a copy‑on‑write copy of that folder inside a sandbox, looks at what every command really changed, and lets nothing reach the real folder until a policy, or you, says so.

Source

Every command runs in the OS sandbox: Seatbelt on macOS, Landlock + seccomp + namespaces on Linux. It can write only to the copy and its own temp folders.

Source

With the default ("balanced") policy: ordinary edits pass; manifests and lockfiles wait for approval; git hooks are denied.

Source

Two tasks can work on the same folder. A file both of them changed merges method by method, and only the parts that were changed twice conflict.

Source