Documentation

Everything you need to stop paying for AI.

Getting Started

Install & Run

# one-shot, no install required
$ npx xswarm-freeloader

This starts an OpenAI-compatible proxy on http://localhost:4011. Point any OpenAI SDK at it.

Global Install

$ npm install -g xswarm-freeloader
$ freeloader start

Set your base URL

# .env
OPENAI_BASE_URL=http://localhost:4011
OPENAI_API_KEY=free # any string

Configuration

Create freeloader.config.json in your project root:

{
"port": 4011,
"budget": {
"monthly_usd": 0 // hard cap at $0
},
"strategy": "free-first",
"trust_tier": "standard",
"providers": ["google", "groq", "together", "ollama"]
}
strategy "free-first" | "cheapest" | "fastest" | "best" How to select models
trust_tier "open" | "standard" | "private" Maximum trust level for routing
budget.monthly_usd number Hard monthly spend cap (0 = free only)
providers string[] Enabled provider adapters
cache boolean Enable prompt-level caching (default: true)

API Reference

Freeloader implements the OpenAI API spec. All existing OpenAI SDK calls work unchanged.

POST /v1/chat/completions Chat completions (streaming supported)
POST /v1/completions Text completions
GET /v1/models List available models
POST /v1/embeddings Text embeddings
GET /health Health check + current routing status
GET /stats Token usage, cost savings, model breakdown

Trust Tiers

Control where your data goes based on sensitivity.

openOpen Source Models

Routes only to fully open-source models (Llama, Mistral, Qwen). Weights are public — no data retention promises needed.

standardStandard Providers

Includes reputable commercial free tiers (Gemini, Mistral API) with standard data processing agreements.

privateAll Providers

No restrictions. Uses any available model including paid commercial APIs when free options are exhausted.

Provider Adapters

Freeloader ships adapters for:

Google Gemini
Groq
Together AI
Mistral
Ollama (local)
LM Studio (local)
OpenRouter
Hugging Face