Runtime protection

Does Guard work with your AI coding tool?

Guard watches what your AI coding tool does: every command it runs, every file it reads, every tool it calls. Risky actions pause for your approval, either inside the tool or in Guard’s local approval center.

6 approve right in the chat. 8 route to the approval center.

Approve right in the chat

These tools prompt you inline. No separate window, no context switch.

6

Codex

In chat

Guard hooks into Codex’s own approval prompts — you approve right in the chat.

Claude Code

In chat

Guard pauses risky actions through Claude’s hooks, right in the terminal.

OpenCode

In chat

MCP approvals stay inline with OpenCode’s allow-once flow. Guard reviews skills and plugins before launch.

Copilot

In chat

Guard prompts through Copilot CLI hooks; the approval center covers what hooks can’t.

Cursor

In chat

Cursor runs its own tool approvals. Guard vets packages, MCP servers, and provenance behind them.

Pi

In chat

A Guard-managed extension reviews prompts and tool calls inline.

Routed to the approval center

For these tools, blocked changes wait in a local review queue you check on your own time.

8

Ready to try it?

One command installs Guard for your tool. Free, open source, takes about 30 seconds.

Install GuardCheck your risk first