Local-first LLM inference

Your hardware.
Any website.
No exposed ports.

Robot Queue connects browser chat requests to Ollama workers running on machines you control — without exposing a single port to the internet.

Get Started Free How it works
🌐
Browser
SDK request
⚙️
Robot Queue
cloud broker
🤖
Ollama Worker
local models
or
Desktop AI App
Claude, ChatGPT…
How it works
Three steps from sign-up to running
01

Create a site

Register your website, configure allowed origins and models, and receive a public site key plus worker credentials.

2 minutes
02

Embed the SDK

Add one script tag to your page and initialize the client with your site key. Call chat.completions.create() like any LLM API.

5 minutes
03

Connect a worker

Run the Ollama CLI on any local machine, or connect your desktop AI app via MCP — Claude, ChatGPT, and others work out of the box.

3 minutes
Simple browser SDK
Works like any LLM API.
index.html
<!-- 1. Include the hosted SDK -->
<script src="https://cdn.robot-queue.robrighter.com/sdk.js"></script>

<script>
// 2. Initialize with your public site key
const client = createRobotQueueClient({
  siteKey: 'rq_site_public_...'
});

// 3. Call chat completions — runs on your local Ollama worker
const result = await client.chat.completions.create({
  messages: [{ role: 'user', content: userMessage }]
});

document.getElementById('output').textContent = result.message.content;
</script>
Two ways to process jobs
Your hardware or your AI app.
🤖

Ollama CLI

Run open-source models locally. The worker script polls silently in the background — pull any Ollama-compatible model and it's ready to process jobs from your queue.

npx tsx worker.ts start --worker-key <key>
  • Llama, Mistral, Phi, Gemma, and more
  • Runs on any machine — laptop, desktop, server
  • No data leaves your network

Desktop AI App via MCP

Connect Claude Desktop, ChatGPT, or any MCP-compatible AI app as a worker. Your AI assistant polls the queue and processes jobs using its own frontier models — no GPU required.

npx tsx mcp-worker.ts --worker-key <key>
  • Claude, ChatGPT, and any MCP client
  • Frontier model quality, zero local setup
  • Switch workers without changing your website
Why Robot Queue
Built for developers who want control.
🔒

Your hardware, your data

Prompts and completions run on machines you own. No third-party model provider ever sees your data.

🌐

No exposed servers

Workers connect outbound to the queue — no inbound firewall rules, no dynamic DNS, no reverse proxies.

Use your GPU for free

Stop paying per-token API fees. If you have a local machine with a GPU and Ollama, you already have an inference server.

🔑

Per-worker keys

Each worker installation gets its own revocable key. Add workers for throughput, remove them instantly if needed.

🔄

Offline-tolerant

Workers reconnect automatically after network interruptions. Job leases expire and are retried if a worker goes silent.

📦

Single JS include

The browser SDK has no dependencies and works in any modern browser. One script tag is all it takes to integrate.

Connect your first site
in under 15 minutes.

Free to use. Runs on your hardware. No credit card required.

Create Free Account