A distributed compute network routing AI workloads to nearby community-operated hardware.
Lower latency. Fair pricing. Data stays local.
Website · App · Docs · API Reference · Blog
Archipelag.io connects people who need AI compute with people who have idle GPUs. Consumers submit inference jobs — LLM chat, image generation — and the platform routes them to the nearest available Island (compute contributor), streaming results back in real time. Islands earn credits for every job they complete.
The API is OpenAI-compatible. Switch one URL and your existing code works.
Currently in open beta through June 2026. Free credits for new accounts.
- OpenAI-compatible API — drop-in replacement, one URL change
- Python SDK —
pip install archipelag→archipelagio-python - JavaScript SDK —
npm install @archipelag/sdk→archipelagio-js - CLI —
island chat "Hello"→island-cli - API reference — interactive docs at api.archipelag.io
Contribute your idle hardware and earn credits. You set the price, the platform matches you with nearby demand. Sandboxed execution means no arbitrary code ever touches your machine — only network-approved, signed containers.
| Repo | Description |
|---|---|
island |
Island agent — runs on contributor machines (Rust) |
island-cli |
CLI for chat, jobs, and account management |
archipelagio-python |
Official Python SDK |
archipelagio-js |
JavaScript/TypeScript SDK with React hooks |
api.archipelag.io |
Interactive API reference (OpenAPI 3.1 + Scalar) |
archipelag.io |
Marketing website |
- Try it — app.archipelag.io
- Build on the API — docs.archipelag.io
- Contribute your GPU — archipelag.io/earn
- Report bugs or ideas — open an issue or email hello@archipelag.io
Named after an archipelago — a network of independent islands working together.