Quickstart

Get up and running with Privane in minutes. Privane connects your high-performance local inference engine to our secure managed tool gateway.

1. Install the CLI

Privane provides a lightweight CLI runtime for managing localized models and spinning up the orchestration layer.

npm install -g privane-cli

2. Start the Local Service

Initialize the background execution engine. This boots a fast, OpenAI-compatible REST server listening locally on localhost:8080.

privane serve

3. Pull a Local Model

Download our highly optimized quantized weights (or drag-and-drop your own standard .gguf files).

privane run gemma-2b

4. Connect the Managed Tool Gateway

Privane’s true differentiator is the ability to securely bridge local reasoning with real-world execution. Register for a developer account on api.privane.dev to unlock our managed cloud connector fabric.

export PRIVANE_API_KEY="your_api_key_here"

The “Wow” Moment: Local Reasoning, Global Execution

Install the orchestrator and tool packages:

npm install @privane/engine @privane/tools

Here is a complete, zero-config script. A local model handles all raw context processing, reasoning, and security decision-making on your hardware, while safely querying GitHub unread notifications through the Privane Managed Tool Gateway:

import { Engine } from '@privane/engine';
import { PrivaneTools } from '@privane/tools';
 
// 1. Initialize local WebGPU/CPU reasoning runtime
const engine = new Engine({ backend: 'webgpu' });
await engine.load('gemma-2b');
 
// 2. Initialize the secure managed tool gateway
const tools = new PrivaneTools({
  apiKey: process.env.PRIVANE_API_KEY
});
 
// 3. Bind the normalized GitHub connector
const github = tools.get("github");
 
// 4. Run the hybrid orchestration loop
const result = await engine.run({
  task: "Check my unread GitHub notifications and summarize the high-priority action items.",
  tools: [github]
});
 
console.log(result.summary);

What just happened?

• Local Sovereignty: The LLM reasoning, output parsing, and execution decisions occurred entirely on your local machine. Your raw private credentials and model context never leaked to centralized clouds.
• Managed Execution: The secure OAuth connection, rate-limiting, and network calls to the GitHub API were offloaded to the pre-configured Privane execution gateway.