Arcloak

Permissionless AI inference on @arc , paid per token in USDC via x402.
Run a node, serve models, earn USDC.

CA: 0x797abf94a828776f082c56e33e5ec432e1d27033

[Arcloak @arcloak]

Arcloak isn’t just for chat.

Anything that can call an OpenAI-style API can use it:
agents
trading systems
research tools
autonomous workflows
apps
image generation
private inference pipelines

Inference becomes another programmable service your software can buy.

[Arcloak @arcloak]

“Private AI” shouldn’t just mean a privacy policy.

Arcloak Private Inference encrypts the prompt before it leaves the client and runs it inside hardware-attested confidential compute.

Then we expose the proof.
Privacy you can verify > privacy you’re asked to trust.

[Arcloak @arcloak]

Arcloak separates the stack cleanly:
API + routing → Arcloak
compute → crowdsourced or confidential nodes
payments → x402
facilitation → Arclight
settlement → USDC on Arc

Modular infra means each layer can improve without changing how developers call the API.

$ARCLOAK makes each layer completely independant

[Arcloak @arcloak]

Have compute sitting idle?
Arcloak turns it into sell-side AI infrastructure.

Run a node.
Serve model requests.
Earn USDC from actual usage.

No points.
No emissions schedule.
No waiting for a token incentive to make the economics work.

[Arcloak @arcloak]

Agents need infrastructure they can actually buy from autonomously.

Arcloak gives them an OpenAI-compatible endpoint where payment is part of the request itself.

No human stopping to create an account.
No manual top-ups.
No billing dashboard.

Just machine-to-machine inference.

[Arcloak @arcloak]

Using AI infra shouldn’t mean juggling accounts, API keys, prepaid credits and subscriptions.

With Arcloak:
call the endpoint
pay per request
get the result
move on

USDC via x402. No account required.

[Arcloak @arcloak]

Arcloak is partnering with @arclightfi to improve how x402 payments settle across the network.

All paid inference already routes through their x402 facilitator, and next we’ll be integrating NEAR account payments as another way for agents and users to pay on $ARCLOAK

Better rails underneath.
Same simple API on top.

[Arcloak @arcloak]

If you have compute, Arcloak turns it into an inference business.

Run a node. Serve model requests. Earn USDC from real usage.

No token emissions. No points farm. Just demand flowing to supply.

[Arcloak @arcloak]

Agents are becoming economic actors.

They’ll need to buy compute, data, APIs and services without humans manually creating accounts for them.

Arcloak is built for that world:

machine requests inference
machine pays for inference
network serves inference.

[Arcloak @arcloak]

Arcloak is basically a crypto-native inference rail:

OpenAI-compatible API
→ encrypted/private or crowdsourced compute
→ x402 payment
→ Arclight facilitation
→ USDC settlement on Arc

Existing AI interface. New payment + infrastructure model.

[Arcloak @arcloak]

Private AI shouldn’t mean “trust us, we don’t look.”

Arcloak’s private tier runs inside hardware-attested confidential VMs on @PhalaNetwork , with prompts encrypted before they leave the client.

And the proof is exposed to the user.

[Arc Hub @ArcHubHB]

Top gainers on Arc last 24h

Which one are you loading up on?👇

$JEANUSDC @Usdcjeanphil
$SCOUT @pokescout_
$ARCLOAK @arcloak
$ARCUS @arcusdotnet
$SOLON @Solonlabs1
$BAGFI @bagfifun
$AI @arcinucto
$ARCT @ArcToolsBackup
$WAVE @WaveonArc
$ARCT @ArcToolsBackup

[Arcloak @arcloak]

New on Arcloak: image generation is live.

Pick gemini-2.5-flash-image in the playground, enter a prompt, sign once, and generate.

~$0.05 per image, paid per request in USDC on Arc via x402.

No account. No API key. No subscription.

https://t.co/Dd3hu33drI

[Arcloak @arcloak]

The future AI stack won’t be 100 different agents holding 100 different API keys.

It’ll be agents finding services and paying for them programmatically.

Arcloak is what that looks like for inference.

[Arcloak @arcloak]

Your agent shouldn’t need an account, API key, prepaid credits and a dashboard login just to use inference.

With $ARCLOAK it calls an OpenAI-compatible endpoint, pays per request in USDC, and keeps moving.

Inference as infrastructure, not a subscription.

[Arcloak @arcloak]

There is a massive amount of idle compute sitting around the world.

There is also a growing amount of AI inference demand.

Arcloak is building the market between them.

Bring compute.
Serve models.
Get paid.

[Arcloak @arcloak]

AI inference is becoming a commodity.

Commodities want open markets.

Arcloak makes inference permissionless on both sides:

Anyone can consume it.
Anyone can supply it.
USDC handles the payment layer.
The market decides the rest.

[Arcloak @arcloak]

Arcloak’s inference stack is deliberately modular:

@PhalaNetwork provides the confidential VM layer for private, hardware-attested inference.
We handles the OpenAI-compatible API, routing, pricing and node coordination.
@arclightfi handles x402 facilitation and settlement.
@arc is where the USDC moves.

Encrypted prompts in.
Attested compute in the middle.
Onchain settlement out.

That’s the stack for $ARCLOAK.

[Arcloak @arcloak]

Want proof your prompt and inference were actually private?

Click Show proof.

Get the attested node, enclave key, quote hash, measurement, verification timestamp + expiry — and links to inspect the node listing and verify the quote yourself.

Private inference shouldn’t be a claim. It should be verifiable.

Only at $ARCLOAK

[Arcloak @arcloak]

Your prompts shouldn’t become someone else’s data.

Arcloak Private Inference encrypts prompts client-side, routes only ciphertext, and decrypts inside a private VM node on Phala with hardware-attested confidential compute.

Inference run on a private node on @PhalaNetwork .
Settlements via @arclightfi .
Private by design.

Pay per prompt using @Alibaba_Qwen private model where your prompt is encrypted using cyphertext and context and prompt is hidden inside a confidential VM.

Not the operator.
Not the orchestrator.

You.

Stay hidden with $ARCLOAK

[Arcloak @arcloak]

Running on our own confidential VM on Phala

[Arcloak @arcloak]

Your prompts shouldn’t become someone else’s data.

Arcloak Private Inference encrypts prompts client-side, routes only ciphertext, and decrypts inside a private VM node on Phala with hardware-attested confidential compute.

Inference run on a private node on @PhalaNetwork .
Settlements via @arclightfi .
Private by design.

Pay per prompt using @Alibaba_Qwen private model where your prompt is encrypted using cyphertext and context and prompt is hidden inside a confidential VM.

Not the operator.
Not the orchestrator.

You.

Stay hidden with $ARCLOAK