# Clyro SDK

A Python SDK and set of tools that add runtime governance to AI agents: cost limits, step limits, loop detection and policy enforcement, plus an MCP wrapper and Claude Code hooks, working offline with local YAML policies.

Clyro wraps a Python agent with clyro.wrap() to add execution tracing, cost tracking, step limits, loop detection and policy enforcement without changing the agent's own logic; if the SDK itself encounters an error it fails open so the wrapped agent keeps running. It works fully offline with local YAML policies and needs no API key, with an optional cloud service for team dashboards, shared policies and session replay.

Beyond the SDK, Clyro includes a policy recommender that inspects an existing agent and suggests the agent type, concerns and policy kits to apply, an MCP wrapper that governs tool calls made through Claude Desktop, Cursor and VS Code, and Claude Code hooks that block destructive commands such as rm -rf or DROP TABLE. It ships adapters for LangGraph, CrewAI, the Claude Agent SDK, the Anthropic SDK and generic agents, and automatically calculates LLM cost for OpenAI and Anthropic models.

It installs with pip, has minimal dependencies, and is licensed under Apache-2.0.

Website: https://clyro.dev/