Laya Guard

Laya Guard is a free local guardrail for AI coding agents that judges every tool call on-device using a small multilingual model.

Laya Guard is a free local guardrail designed to protect AI coding agents by checking every tool call locally on your machine. It uses a regex prefilter for known-dangerous patterns and a small multilingual model named Laya for everything else, resulting in allow, review, or block verdicts. The system measures latency at roughly 85 to 90 milliseconds per judgment on a laptop CPU and ensures that safe commands return instantly without invoking the model. It includes a local dashboard application that operates entirely on your machine to monitor and manage protected agents. Laya Guard supports integrations with eleven coding agents including Claude Code, Codex CLI, Cursor, Gemini CLI, Qwen Code, Windsurf, Copilot CLI, Cline, OpenCode, Kilo Code, and Amp.

Laya Guard at a glance

Best for: Software engineer
Runs on: macOS, Linux, API, CLI, self-hosted, open source
Pricing: Free
Released: October 2026
Category: AI coding tool
Website: github.com/scs0209/laya-guard

Key features

Judge tool calls locally in ~0.1s
Combine regex prefilters and multilingual model judgment
Provide a local web dashboard for agent management
Support eleven AI coding agents out of the box
Run completely offline with no network loading
Check package.json scripts against regex policies
Back up and write configuration files atomically

Use it for

Guard Claude Code tool executions
Protect Cursor before shell execution
Monitor AI coding agent shell commands locally
Review policy decisions and blocked calls

Alternatives to Laya Guard

AgentGuard — Get a security and quality assessment of AI skills in seconds without an account.
Secret Guard — Prevents sensitive credentials from reaching third-party AI vendor servers.
Repo Guards — It prevents AI agents from performing destructive actions across multiple repositories.
Guardrails AI — It detects and mitigates specific risks and generates structured data from LLMs.
Coucou — It provides open-source, real-time agent monitoring and approvals directly from the screen notch.
Stop That Shit — Enforces task boundaries to stop unnecessary AI agent actions.
Agent Usage Guard — Stops runaway usage windows and excessive subagent fan-out.
Yara L Skill — Teaches coding agents to generate YARA-L 2.0 rules and includes an offline linter.
Destructive Command Guard — It prevents AI agents from accidentally running commands that destroy work.
OpenAPPA — Tracks data flows algebraically to prevent prompt injection without breaking agents.
Spring AI Privacy Guardrails — Protects sensitive data and secures tool usage within Spring AI applications.
Claude Code Mistake Guards — Stops repeated agent mistakes mechanically with tests that prove they block.

All Laya Guard alternatives