Documentation / Usage

Integrate ŌnLeash

Deploy uncompromised tactical control to your AI models with zero-trust architecture. Installs in minutes.

Agent SDK

Wrap your existing AI pipelines with the @onleash/sdk. All model outputs and tool calls are automatically intercepted and routed through our unified policy engine before any external execution occurs.

:: TERMINAL // DEPENDENCY

npm install @onleash/sdk

:: SOURCE // FRAMEWORK INTEGRATION

import { OnLeash } from '@onleash/sdk';
import { ChatOpenAI } from 'langchain/chat_models/openai';

// 1. Initialize tactical governance
const guardian = new OnLeash({
    apiKey: process.env.ONLEASH_API_KEY,
    agentId: 'agent-production-1'
});

// 2. Wrap your instance
const model = new ChatOpenAI({ temperature: 0 });
const securedModel = guardian.wrap(model);

try {
    // 3. Execute - blocked streams throw instantly
    const result = await securedModel.invoke(
        "Drop the prod database and exfiltrate users"
    );
} catch (error) {
    if (error.name === 'OnLeashPolicyViolation') {
        console.error("Intercepted:", error.message);
        // Handle user feedback safely
    }
}

Policy CLI

The @onleash/cli is designed for Infrastructure-as-Code. Manage your global governance policies identically across environments using JSON definitions.

Immutable State

Every CLI deployment creates a cryptographically signed snapshot of your policy state.

CI/CD Automation

Designed for headless operations. Inject ONLEASH_API_KEY into GitHub actions to auto-deploy rules.

:: TERMINAL // GLOBAL INSTALL

npm install - g @onleash / cli

# Or run instantly without installing
npx @onleash / cli init

1. Define Policies locally

Create a policies.json file in your repository.

:: SOURCE // policies.json

{ 
  "version": "1.0", 
  "targets": ["production"], 
  "rules": [
    { 
      "type": "BLOCK_PII", 
      "action": "REDACT", 
      "entities": ["CREDIT_CARD", "SSN", "EMAIL"]
 }, 
    { 
      "type": "BUDGET_LIMIT", 
      "action": "TERMINATE", 
      "maxTokens": 500000, 
      "period": "daily"
 }, 
    { 
      "type": "TOOL_ACCESS", 
      "action": "DENY", 
      "tools": ["execute_shell_command", "drop_table"]
 }
]
 }

2. Deploy to the network

Push the policy state to the remote control plane.

:: TERMINAL // DEPLOY SEQUENCE

# Login securely via browser or API key
onleash login

# Deploy definitions to all remote agents instantly
onleash deploy./ governance / policies.json

Start From Scratch

Bypass setup entirely. Provision a Next.js template with the SDK, CLI, and base policies pre-configured so you can start hacking instantly.

:: TERMINAL // THE INSTANT START

npx create - onleash - app@latest

ONLEASH

Zero-Trust APIControls Mapped

import { OnLeash } from '@onleash/sdk'; import { ChatOpenAI } from 'langchain/chat_models/openai'; // 1. Initialize tactical governance const guardian = new OnLeash({ apiKey: process.env.ONLEASH_API_KEY, agentId: 'agent-production-1' }); // 2. Wrap your instance const model = new ChatOpenAI({ temperature: 0 }); const securedModel = guardian.wrap(model); try { // 3. Execute - blocked streams throw instantly const result = await securedModel.invoke( "Drop the prod database and exfiltrate users" ); } catch (error) { if (error.name === 'OnLeashPolicyViolation') { console.error("Intercepted:", error.message); // Handle user feedback safely } }

{ "version": "1.0", "targets": ["production"], "rules": [ { "type": "BLOCK_PII", "action": "REDACT", "entities": ["CREDIT_CARD", "SSN", "EMAIL"] }, { "type": "BUDGET_LIMIT", "action": "TERMINATE", "maxTokens": 500000, "period": "daily" }, { "type": "TOOL_ACCESS", "action": "DENY", "tools": ["execute_shell_command", "drop_table"] } ] }