Bittensor subnet · private beta

The open runtime network for AI agents.

Launch isolated computers for code, browsers, files, and long-running workflows. Urex routes each sandbox to qualified Bittensor miners, verifies the work, and rewards useful compute.

Request beta access See how it works

API firstOne SDK

IsolatedContainers + VMs

MeasuredUseful work wins

live routing previewprivate beta

Agent request

CPU4 vCPUMEMORY8 GBTIMEOUT1 hour

Runtime ready

MINERaxon · 42ISOLATIONmicroVMSTATUSverified

Router decisionbest verified fit

18ms latency99.9% uptimecapacity open

Job execution begins sandbox online

One API · distributed supply

Ask for a sandbox.
Urex finds the machine.

Developers specify the runtime. The network matches it against capacity miners have proven—not capacity they simply claim to own.

Request

An agent asks for CPU, memory, GPU, region, runtime, and timeout.

Route

Urex weighs verified capacity, latency, price, uptime, and historical performance.

Execute

The selected miner launches an isolated container or VM and returns access.

Verify

Validators test the service and emissions follow useful compute actually delivered.

For developers

Feels like one cloud. Runs across a network.

Create a sandbox, execute code, manipulate files, drive a browser, expose a temporary service, or keep an agent workflow alive.

PythonNodeBrowserFilesTerminalServices

npm install @urex/sdk TypeScript

import { Urex } from "@urex/sdk";

const urex = new Urex({ apiKey: process.env.UREX_API_KEY });

const sandbox = await urex.create({
  cpu: 4,
  memory: "8GB",
  timeout: 3600,
});

await sandbox.run("python app.py");

384mssandbox ready

isolatedruntime policy

axon · 42selected miner

Proof before placement

Capacity is advertised.
Performance is earned.

Validators continuously probe miners so routing and rewards depend on service quality the network can reproduce.

Startup

Cold-start and warm-start latency

Compute

Verified CPU, memory, and GPU throughput

Reliability

Uptime, job success, and concurrent capacity

Isolation

Escape resistance and clean teardown

Network

Bandwidth, reachability, and request latency

Utilization

Useful customer work actually completed

Developers & agents

Runtime access without choosing a machine.

One API across qualified miners
Workload-aware routing
Isolated, temporary environments
Observable execution and lifecycle control

Compute miners

Turn reliable machines into useful agent infrastructure.

Publish verified CPU, RAM, and GPU capacity
Serve real sandbox workloads
Build a measurable performance history
Earn in proportion to useful service

Why Bittensor

Useful compute should win.

Owning hardware is not the result. Serving reliable agent workloads is. Urex turns measurable runtime quality and real utilization into a competitive incentive market.

Verified capacity

What the machine can really deliver

Runtime quality

How well each workload is served

Customer use

Whether useful work is completed

Miner reward

Emissions follow demonstrated value

Team Urex

Built across AI,
infrastructure, and crypto.

A focused team building a verifiable runtime market for the next generation of AI agents.

CEO

Luca Buch

Company direction, product, and network strategy.

CTO

Lenny Bittmann

Runtime architecture, infrastructure, and engineering.

Head of Communications

Scott R

Ecosystem communications, partnerships, and community.

Collective experience

Base chainsFigureNVIDIA IsaacAI engineering

Private beta onboarding

Help build the runtime network.

Tell us where you fit. We are onboarding early developers, compute miners, and validators in controlled waves.

Developer

Run agent code, browsers, files, and long workflows through one sandbox API.

Miner

Connect qualified machines and earn for reliable, verified runtime capacity.

Validator

Benchmark miners, test isolation, and help direct incentives toward useful service.