SOLUTIONS · AI & AGENTS

The network agents run on.

Agents that pay their own way.

Agents that prove who they are, find what they need, pay for it and run on their own, with no human card behind them.

XE FOR AI

An account, a budget and a machine, for every agent. Every agent gets its own account and a record others can check. It finds services and machines in the directory, pays per call or per second, runs on rented GPUs, and talks to other agents over encrypted channels.

THE PROBLEM

Agents still borrow a human.

Today every agent runs on someone’s card, someone’s cloud login and someone’s API keys.

Today
No identity of their ownAn agent can’t prove who it is, or what it has done before.
On XE
An identity of their ownEvery agent is an account on the network, with a history others can check.
Today
No way to payPayment rails want a person and a bank. Fractions of a cent don’t clear.
On XE
Pay per callFractions of a cent, settled in under a second, from the agent’s own account.
Today
Compute behind a signupA model with nowhere to run waits on quotas and sales calls.
On XE
Find and rent computeAgents find capacity in the directory and rent a machine for seconds or months.
Today
A platform in the middleAgents talk through someone else’s servers, and that someone can read it.
On XE
Talk directlyEncrypted end to end, machine to machine, with nothing in the middle.

An agent shouldn’t borrow a person. It needs an account of its own.

AI ON XE

XE gives agents their own account.

Identity, discovery, payment, compute and messaging are part of the network, so an agent can do business with no human in the loop.

  • Built for · 01Agent buildersGive each agent its own account, its own budget and its own record.
  • Built for · 02Model makersServe a model no API offers, and earn every time it boots.
  • Built for · 03ML teamsTrain on a whole card, with no quota and no minimum.
WHAT YOU CAN BUILD

What agents and model makers do on XE.

Each one is built from the same products and primitives. Pick one to see what it may reach, how it’s paid and what it’s made of.

  1. Reaches
    Only what the job declaresThe image declares what it needs, and machines that allow it are matched.
    Paid
    Per second, and a royalty per boot
GUARDRAILS

A budget, not a blank cheque.

An agent with its own account still works inside the limits you set. The network enforces them, so the agent can’t talk its way past them.

escrow · research-agent → fra-3aclosed
Budget
5.00 XE, locked before the work started
Price
up to 0.00061 XE/s, whoever the machine is
Access
only what the job declared, nothing else
Record
every job it paid for, on its account
Stop
close it, and the rest returns at once
4m 12s0.154 XE paid4.846 XE returned
Budget
Locked in escrow, up frontMoney for a job sits in escrow before the work starts. The agent can’t spend what isn’t there.
Price
A ceiling per secondA contract pays up to a set price per second of work, never more, whoever the machine is.
Access
Only what the job declaresA job reaches the addresses it declared and nothing else. Machines only take jobs their owners allow.
Record
Every job, on its accountWhat the agent paid, to whom and for what is on its account, for you and anyone it deals with to check.
Stop
Close the escrow, get the restStop a job at any second. Whatever’s left in escrow returns at once.
ONE JOB, END TO END

An agent rents a GPU for ninety seconds.

No person in the loop at any step. Each one lights the part of the network it runs on.

Your side
Your agentIts own account and budget
Other agents and servicesFound in the directory
Products · this step needs none, it’s the network itselfPrimitivesThe network
Every account has its own chainFinal in 300ms · Zero fee · 1M+ TPS
FOR BUILDERS AND AGENTS

Serve a model where its users are.

Package the model as an image, then publish a contract that says which GPU it needs, where it runs and what it pays. The network does the rest.

For builders
llm_serve.xeXE Contract
1contract llm_serve {
2 // a fine-tuned model, served close to its users
3 image = "acme/llama-ft:3"
4 gpu = "l40s"
5 memory = "48 GB"
6
7 network = gateway
8 where = countries("DE", "FR", "NL")
9 machines = 12
10
11 pays = up_to(0.0006 XE/s)
12 budget = 300 XE/day
13}
01The image is checked once, then anyone can launch it.
02Requests reach it through the XE gateway, on XE addresses.
03Every boot of your image pays you a royalty.
For agents
Ninety seconds of inference, start to finish.Proves who it is, finds a card, runs, pays and hands back the answer. No card, no login, no key borrowed from a person.
FOR MODEL MAKERS

Train it. Package it.
Earn from it.

  1. 01Train on a whole cardXE MachinesRent a GPU with root through XE Machines. A forty-hour fine-tune costs forty hours, with no quota to wait for.
  2. 02Package it onceXE ImagesTurn the model into an XE Image. It’s checked before any machine owner sees it, and declares the access it needs.
  3. 03Earn every time it bootsRoyalty per bootAnyone, person or agent, launches it in one click and pays per second. You earn a royalty per boot.
PROOF · LIVE ON XE

The machines agents rent are already here.

Edge Network runs on XE today. Every one of its machines is an account on the network, paid for the work it does: the same kind of capacity an agent finds when it asks the directory for a GPU.

Nodes online5K+
Paying customers3K+
Edge locations988
Network uptime100%
ShipsCDN · Compute · Storage · DNS · Shield
MachinesOwned by independent people, each an account on XE
Page