# @kollguard/agent-sdk

> Gate Claude Agent SDK tool calls through KollGuard Agent Watch: a PreToolUse hook (and canUseTool fallback) that allows, denies, or waits for a human approval. Sends tool and argument names only, never values.

Latest version **0.2.0** (published 2026-11-19) · MIT license · 0 weekly downloads

_The full health report (score, vulnerabilities, README, dependency tree, version history) is being computed; this document will grow once it lands._

## Install

```sh
npm install @kollguard/agent-sdk
pnpm add @kollguard/agent-sdk
yarn add @kollguard/agent-sdk
bun add @kollguard/agent-sdk
```

## Facts

| | |
|---|---|
| Version | 0.2.0 |
| Published | 2026-11-19 |
| First published | 2026-11-19 |
| Weekly downloads | 0 |
| License | MIT |
| TypeScript types | bundled |
| Module format | ESM + CommonJS |
| Node | >=18.17 |
| Dependencies | 2 |
| Unpacked size | 22.6 KB |
| Author | Kollitech |
| Maintainers | kollguard |
| Keywords | claude-agent-sdk, claude, anthropic, hooks, pretooluse, canusetool, kollguard, agent-governance, human-in-the-loop, guardrails |

## Links

- npm: https://www.npmjs.com/package/@kollguard/agent-sdk
- Homepage: https://kollguard.com/agent-authorization#claude-code
- npm.io page: https://npm.io/package/@kollguard/agent-sdk

---
_Source: https://npm.io/package/@kollguard/agent-sdk · Machine-readable twin of the npm.io package page. Health data is recomputed on every publish._
