Lelu Engine v1.0 is live

The authorization layer
for AI Agents.

Control what your agents can do. Lelu brings confidence-aware gating, interactive human-review, and tamper-proof auditing to your autonomous AI workflows.

Get Started View on GitHub

agent.ts

import { lelu } from '@lelu/sdk';
import { OpenAI } from 'openai';

// Wrap your agent tools securely
const agent = openai.beta.assistants.create({
  instructions: "You are a helpful billing assistant.",
  tools: [
    lelu.secureTool({
      name: "issue_refund",
      policy: "billing_rules.yaml",
      requiresConfidence: 0.90,
    })
  ]
});

// If confidence < 0.90, the agent is securely paused and
// routed to the Lelu Human Review dashboard.

Security infrastructure for your agents.

A comprehensive suite to monitor, control, and audit autonomous workflows.

Policy-Driven Decisions

Lelu applies Rego policies at action time so every tool call can be evaluated, gated, and audited before hitting your APIs.

Human-in-the-Loop

High-risk AI actions (like issuing refunds or dropping tables) are automatically paused and routed to human reviewers in a beautiful UI.

Tamper-Proof Audit Trail

Every allow, deny, and escalation is captured with confidence context, trace IDs, and decision metadata for absolute observability.

Multi-Agent Delegation