Welcome

Klent sits between your agent and the tools or APIs it calls. Before an action executes, Klent evaluates it against your policies and decides whether to allow, deny, or modify it. Every decision and outcome is recorded on a per-execution timeline you can query, export, and audit.

This documentation is organized in four sections:

• Getting started — from npm install to a blocked action in five minutes.
• SDK reference — the TypeScript and Python clients, every method, every option.
• Concepts — what an execution is, how the policy engine evaluates, how the pieces fit.
• API reference — endpoint-by-endpoint docs (coming soon; for now the TypeScript SDK mirrors the HTTP surface).

What Klent is not

Klent is not an agent framework. It does not provide LLM access, prompt tooling, or orchestration. It is a thin, opinionated layer that works with any agent stack — OpenAI Agents SDK, Anthropic tool use, LangChain, Mastra, LlamaIndex, or your own loop.

What you need before starting

• A Klent project and API key. Sign up at the dashboard — a project and key are provisioned on first sign-in.
• Node.js ≥ 20 for the TypeScript SDK, or Python ≥ 3.10 for the Python SDK.
• An agent that calls tools. Klent is most useful the moment you give an LLM the ability to take real actions.