The open inference network.

Repogen routes AI workloads to a distributed network of idle Macs. Developers get OpenAI-compatible access to frontier models at half the cost. Mac owners earn USDC from compute they already own. No cloud middlemen.

ARENA

BECOME A PROVIDER

50%

lower pricing

95%

to providers

USDC

on Base L2

Inference infrastructure without the infrastructure.

Traditional inference requires provisioning GPUs, managing capacity, and paying for idle time. Repogen abstracts this entirely. Your API calls route to a distributed network of Apple Silicon machines running MLX-optimized models. You pay per token. Providers earn per token. The network handles everything in between.

ROUTING

Requests match to providers based on model availability, latency, and capacity. The coordinator handles load balancing and failover automatically.

INFERENCE

Providers run MLX on Apple Silicon. Unified memory architecture enables large models on consumer hardware. M2 Ultra can serve 70B parameter models locally.

SETTLEMENT

Usage metering happens at the coordinator. Settlement in USDC on Base. Sub-cent transaction costs make per-request micropayments viable.

For developers.

OpenAI-compatible API. Change your base URL and keep your existing SDK, request format, and streaming code. Access Llama, Qwen, DeepSeek, and other open models at roughly half the cost of centralized providers.

—Drop-in replacement for OpenAI SDK
—Streaming responses via SSE
—Function calling and tool use
—Pay only for tokens consumed

example.py

from openai import OpenAI

client = OpenAI(
    base_url="https://api.repogen.xyz/v1",
    api_key="rg_..."
)

response = client.chat.completions.create(
    model="gpt-oss-20b",  # or gemma-4-26b
    messages=[
        {"role": "user", "content": "Hello"}
    ],
    stream=True
)

for chunk in response:
    print(chunk.choices[0].delta.content)

For Mac owners.

Turn idle compute into income. Install the provider software, configure your availability, and start serving inference requests. The network matches your hardware with demand. You keep 95% of earnings.

—Works on any M1/M2/M3/M4 Mac
—Set your own availability schedule
—Automatic model downloads from HuggingFace
—Earnings settle directly to your wallet

Terminal

$ curl -fsSL https://repogen.xyz/install | bash
$ repogen start

Provider registered: M4 Max · 128GB
Waiting for inference requests...

EARNINGS

Earnings vary based on your hardware specs, availability, and network demand. Higher memory capacity supports larger models with better rates.

Public alpha is live.

Connect your wallet, deposit USDC on Base, and start building. Providers can install and begin earning immediately.

OPEN DASHBOARD

READ DOCS