# Jadegate Sdk

> Protect AI agent tool calls with 2 lines of Python code. Auto-hooks into OpenAI/Anthropic SDK tool calls with zero configuration.

- Skill: `jadegate/jadegate-sdk` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jadegate/jadegate-sdk`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jadegate/jadegate-sdk/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: jadegate (https://skillmd.com/u/jadegate)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jadegate/jadegate-sdk

---


# JadeGate Python SDK

Protect AI agent tool calls with 2 lines of Python code. Auto-hooks into OpenAI/Anthropic SDK tool calls with zero configuration.

## When to Use

- User is building a Python AI agent and wants security
- User asks about protecting tool calls in code
- User wants to add security to their MCP client programmatically

## Usage

```python
import jadegate
session = jadegate.activate()          # All SDK tool calls now protected

# Optional: custom policy
jadegate.activate(policy="my-policy.json")
jadegate.activate(policy={"max_calls_per_minute": 30})

# Deactivate when done
jadegate.deactivate()
```

## Features

- Zero external dependencies
- Auto-hooks into OpenAI and Anthropic SDK tool calls
- Fully offline — no cloud, no telemetry
- Python 3.8+ compatible