Interlinked CLI

interlinked-cli

The harness for your harness. A local guard layer for AI coding agents —
integrates with Claude Code and Codex (supported), plus experimental Cursor,
Copilot CLI, Gemini CLI, OpenCode, and Pi adapters; evaluates delivered tool
hooks against deterministic rules; blocks the dangerous ones in
milliseconds; keeps a local activity log you can grep.

Install in one line

Node.js 22+ on macOS, Linux, or WSL. Pick one:

Then, inside the repo you want guarded:

Restart your agent so it loads the installed hooks and skills. The guard checks
events delivered by that client's supported hook integration. Use
interlinked harness capabilities --json to inspect installation and observed
coverage. These instructions install this repository directly; full details are in
Install From Source.

The exit ramp, up front: interlinked disable non-destructively stands the guard down.
interlinked disable --uninstall removes an enable/wizard installation; the narrower
interlinked uninstall-hooks removes only hooks installed through install-hooks —
both preserve a managed bridge if somebody changed its bytes.
The local guard requires no remote service. Runtime state lives under
.interlinked/; installation also writes the selected providers' settings,
managed bridges, and skill directories. Removal is explicit. Try it on a
throwaway repo first if you like; the
whole setup is interlinked (the wizard), and its matching teardown is
interlinked disable --uninstall.

Why this matters for responsible AI at scale

Once an AI agent can write code, run shell commands, and install dependencies
on its own, the question for any organization stops being *"can it help?"* and
becomes *"how do we prove every controllable action a configured runner exposed
was checked against policy before it ran —…