[ AETRON Constructor ]

Build and integrate your own AI services on AETRON — decentralised infrastructure running on a blockchain

Run and fine-tune models on AETRON’s decentralised network of GPU miners. A deployed service is an AETRON neuronet.

Create a neuronet →Model catalogue

What is the Constructor?

The Constructor is a console for launching neuronets on the AETRON network without writing any code. You choose the model, the tasks and how you want to earn. Everything that touches the blockchain is handled for you.

Chain keys are held on the platform side. It signs transactions for you and pays the registration and the activation stake — the minimum needed to launch — so you need neither a browser wallet nor a seed phrase.

Custody: the server holds the keys and signs; you log in with email.

No-code: a form in the browser instead of contracts and a CLI.

Hosted: we run the API gateway, and users call your task directly.

The request path: from the owner through the constructor and the gateway to the miners and back, with the reward returning to the ownerYour AI product[ API ]constructorGatewayroutes requestsMinersinference & trainingAET rewardsowner's shareThe request path: from the owner through the constructor and the gateway to the miners and back, with the reward returning to the ownerYour AI product[ API ]constructorGatewayroutes requestsMinersinference & trainingAET rewardsowner's share
89
models
Ready registry
8
text · images · video · sound · vision · embeddings
Modalities
2–5
minutes to launch
Deployment
AET
settlement & rewards
Network token

One POST — and a verification receipt

The request body is a single input field with a key in the header. Every answer has a checkable status: a sample of requests is recomputed by independent miners, and the vote is visible by request id.

request
$ curl -X POST \
  https://gateway-testnet.aetron.ai/n/7/v1/0/infer \
  -H "x-api-key: ak_live_..." \
  -d '{"input": "Explain staking in one line"}'

{
  "id": "9f2c…e41a",
  "output": "Staking locks AET to raise capacity…"
}
answer verification
$ curl https://gateway-testnet.aetron.ai/n/7/v1/requests/9f2c…e41a \
  -H "x-api-key: ak_live_..."

{
  "id": "9f2c…e41a",
  "verification": {
    "status": "verified",
    "votes": 3,
    "quorum": 3
  }
}

What you can deploy

A sample of requests is re-run by other miners — 7% by default. Every request carries its verification status.

One neuronet holds several tasks of different types — chat, images, vision, training, music — each with its own endpoint, and the network’s miners execute themChat & text63 language modelsImagesFLUX.2, Qwen-ImageVisionVLM & OCRTrainingLoRA & SFT on-chainMusicACE-Step, MiniMaxPrivacyencryption in transityour neuronetcomputed by minersOne neuronet holds several tasks of different types — chat, images, vision, training, music — each with its own endpoint, and the network’s miners execute themyour neuronetChat & text63 language modelsImagesFLUX.2, Qwen-ImageVisionVLM & OCRTrainingLoRA & SFT on-chainMusicACE-Step, MiniMaxPrivacyencryption in transitcomputed by miners

What you can build on this

From a plain API endpoint to a model trained for one job.

01

Chatbot / LLM API

An open LLM behind an OpenAI- and Anthropic-compatible endpoint: an SDK connects by switching base_url.

Example · Qwen chat for a support product

02

Content generation

Images and text for creative tools, at whatever volume you need.

Example · an image endpoint on FLUX.2

03

AI agents

A backend for autonomous agents. It stays up as long as requests keep coming, and answers can be verified.

Example · Tool-using agent runtime

04

Fine-tuning

Take a base model and train it for your own domain, right on the network.

Example · LoRA on a domain dataset

How owners earn

Inference is free for your users — nobody takes a cut from them. Hosting is paid by subscription.

01

Emission share

A neuronet earns network emission for useful work. The owner gets up to 10%: the grant programme sets the share at launch, and it can be lowered but never raised.

up to 10% to the owner
02

Hosted API

Access is handed out with API keys: a key can be scoped to one task, a daily budget and a list of addresses. Collecting payments from your users is not supported yet.

Per-task keys
03

Staking capacity

The more AET staked, the more requests the neuronet holds. Stake is added from the AETRON extension or the desktop client — the web console holds no wallet.

Capacity = stake

Why the Constructor

01

No code, no infrastructure

You don't write contracts, run a node or fight with a CLI. It all happens through a form in the browser.

02

The platform pays

The platform pays registration and the activation stake out of its grants. You do not need to fund a wallet of your own to start.

03

A ready registry

Models in the registry are already vetted, with their hashes and specs filled in. No manual manifest work.

Who's in the economy

Three sides share the reward of every neuronet.

01

Owners

Launch neuronets and decide the tasks and the terms.

Rewards · up to 10% of neuronet emission

02

Stakers

Put up AET and raise the neuronet's capacity.

Rewards · a share of neuronet rewards

03

Miners

Lend their GPUs for inference and training.

Rewards · emission for verified work

How to launch

Five steps from sign-up to a working neuronet.

01

Sign in

Register with an email. No wallet needed.

Needs · Email

02

Pick a model

Pick a model in the catalogue — its parameters fill themselves in. The task is added after the neuronet launches.

Needs · Registry

03

Set the terms

Set the name and the execution mode. The starting stake and capacity come from the grant programme.

Needs · A few clicks

04

Deploy

The platform signs the transactions and pays for the launch.

Needs · AETRON Grants pays

05

Get your API key

Issue a key for the task and point your app at it.

Needs · Per-task keys

Build your neuronet

Launching takes a few minutes. No wallet and no servers of your own are needed to launch.

Create a neuronet →Browse the registry