Now in Early Access

One API.
Every Multimodal Model.

Run open-source AI on a decentralized infrastructure — cheaper, faster, safer, always on.

model: flux-1-schnell · region: auto · ~0.9s per image

Free $1.00 credit on signup. No credit card required.

6
Models
4
Regions
6
Modalities
99.9%
Uptime Target
Base URLhttps://api.ecopool.ai/v1

How it works

Go from zero to production in minutes

01

Get Your API Key

Sign up and grab your API key from the dashboard. Free $1 credit included — no credit card required.

02

Swap One Line

Point your OpenAI SDK to api.ecopool.ai. Same endpoints, same format. Your existing code just works.

03

Ship to Production

Multi-region routing, automatic failover, and adaptive scaling handle the rest. Focus on building, not infrastructure.

Why ECOPOOL

Purpose-built infrastructure for production AI

Shorter Leadtime

Every request is routed to the nearest node in our global edge network. Minimized hops, optimized paths — your inference runs closer to your users with lower latency.

Better Uptime

Decentralized compute with no single point of failure. Multi-region by design — if one node goes down, traffic automatically shifts to the next healthy region.

Lower Cost

Vertically integrated hardware purpose-built for multimodal workloads — not repurposed LLM chips. We pass the efficiency gains directly to you.

Agent Ready

Drop-in MCP-compatible API that works out of the box with autonomous agent frameworks. Build, orchestrate, and scale AI agents without infrastructure headaches.

What $1 gets you

Simple, transparent pricing. Pay only for what you use.

500K
LLM tokens
100K
Vision tokens
50
Images
10
Video clips
60min
Transcription
$0
Platform fee

Everything you need to deploy AI

From instant API access to dedicated model endpoints — pick the plan that fits your workload.

Inference API

Pay per token across all platform models. No minimum commitment.

From $0.03/1M tokens
  • 6 platform models
  • OpenAI-compatible
  • Multi-region routing
  • $1 free credit
Popular

Dedicated Inference

Your own model endpoint with multi-region failover. Deploy any HuggingFace model.

Per GPU-hour · varies by GPU
  • Multi-region failover
  • Any HuggingFace model
  • Up to 96GB GPU memory
  • vLLM or custom container