All week we have been showing you what Private AI inference looks like in practice.

Tomorrow, you will see who makes it possible.

Cluster Protocol is now Top 5 on Base.

300K+ wallets and 25K+ dApps deployed through CodeXero, now ranked among the most popular apps on the network.

The orchestration layer for autonomous workflows, live where builders already are.

There's never been a better time to Build on Base.

There's a reason you don't paste your bank statements into a chatbot.

Now you don't have to think twice.

Cluster x Phala: private inference on the models you already use, DeepSeek, Qwen, GLM, MiniMax, running sealed inside TEE hardware. Your prompt is invisible to everyone in the path. Including us.

Every call returns a signed attestation, proof your data never left the enclave.

You don't have to take our word for it. The hardware proves it.

Full breakdown ⬇️

Deep dive on verifiable private AI inference.

GLM 5.1 is now live on Cluster Protocol's Private AI inference infrastructure, powered by Phala TEE.

ZhipuAI's most advanced model, running inside a hardware-level Trusted Execution Environment. Enterprise-grade reasoning with hardware-level privacy guarantees.

Your inputs never leave the secure enclave.

Try it at http://hub.clusterprotocol.ai/models

Your AI agent just paid for something. you did not approve it.

Cluster roundtable #6: what happens when AI agents start spending money on their own?

Agents with wallets are buying compute, data, and settling transactions with no human in the loop. we put the people building those rails in one room to argue about what breaks and who gets automated.

🎙️ hosted by YJ, chief, Cluster Protocol

ft.
~ Phong Nguyen Dang, head of BD, InFlow
~ George, strategic partnerships, 0G Labs
~ Fehmi, CSO, VS1
~ Parul Gujral, founder, Snowball Money
~ David, cofounder, Inflectiv AI
~ Shivang Kamboj, head of growth, Lighthouse
~ Mukund Rawat, sr. backend and smart contracts dev, NodeOps
~ Arjun Kalsy, chief of growth, Covalent
~ Michael Stephen, community lead, Pulsar Money
~ Myo, CMO, GamePad
~ Eniitan, community manager, manadia

🗓 Fri, June 19 | 2:30 PM UTC
🔗 Space link ⬇️

https://twitter.com/i/spaces/1nKOLLWjMgyGR

Set your reminders now!

MiniMax M2.5 is now live on Cluster Protocol's Private AI inference infrastructure, powered by Phala TEE.

Built for long context. MiniMax M2.5 handles extended inputs inside the TEE without compromising privacy at any point in the process.

Long documents, complex workflows, confidential data, all processed privately from start to finish.

Try it at hub.clusterprotocol.ai/models

Will the next billion-dollar product be built by someone who can't code?

Let's find out at our next roundtable session.

We are bringing together builders from across AI, data, and infra to break down what it actually takes to ship something big today.

🎙️ Hosted by: Anjali

ft.

~ Shivang Kamboj, Head of Growth, Lighthouse
~ Arjun Kalsy, Chief of Growth, Covalent
~ Azibi, Community Manager, DataVLT
~ Mukund Rawat, Sr. Backend & Smart Contracts Dev, NodeOps
~ Zakaria Tchikadze, COO, LLMTune.io
~ Myo, CMO, GamePad
~ Advait, CSO, OpenGradient

🗓 Friday, June 26, 2:30 PM UTC

🔗 Space link below ⬇️

https://twitter.com/i/spaces/1PKqrrkzANdGb

Don't miss.

Three open protocols are turning isolated AI agents into a working economy: MCP to use tools, A2A to talk, x402 to settle.

Nobody owns the stack.

We broke down why that's the whole point, and where the value actually pools.

Full breakdown here at the Cluster blogs

Calling a model "open source" because you can use it is like calling a car open source because you can drive it.

The real question is whether you can open the hood.

Azibi from DataVLT on what openness at the model layer actually requires.

Qwen3.5 397B-A17B is now live on Cluster Protocol's Private AI inference infrastructure, powered by Phala TEE.

397 billion parameters. Running privately inside a hardware-level Trusted Execution Environment.

The largest model available on Cluster Protocol's Private AI infrastructure, for teams that need maximum capability without sacrificing data privacy.

Try it at hub.clusterprotocol.ai/models

Cluster x @MindNetwork_xyz

Every payment an AI agent makes through an open standard is public by default. Sender, receiver, token, amount, permanently onchain.

Fine for a hobby agent. A non-starter for a business running agents at scale.

Mind Network is fixing that with Cluster.

Cluster's agents already pay per call through x402.

Mind Network's x402z wraps that same payment in confidentiality: it still settles and stays verifiable, but the sender, receiver, and amount are encrypted.

Full picture at Cluster Blogs.

Built on Cluster. Funded through x402. Private through x402z.

Cluster x NeoSoul

NeoSoul runs the credibility layer for the agent-to-agent economy: an agent school, a skill market, and an agent capital market that test AI judgments against settled real-world outcomes and turn them into computable reputation.

Cluster becomes the build, payment, and data layer underneath it.

An agent built on CodeXero carries a NeoSoul track record from the moment it deploys. It pays its own way through x402, and NeoSoul scores it creditworthy for every call it settles.

The reputation it earns becomes a tradable data class on Cluster's marketplace.

Read at Cluster Blogs on how the pieces connect.

Who owns the oil of AI?

Everyone talks about AI models.

Fewer people talk about where the data lives, who controls access to it, and whether it survives long enough to matter.

Be ready for Founder’s Cut by YJ, Chief, Cluster Protocol as he sits down with Nandit Mehra, Founder, Lighthouse to discuss all of it.

8th July, 9:30 AM UTC.

Live on X.

Two lives of crypto: the one people trade, and the one people actually use.

We're sitting down with @Cwallet_Exchange, a crypto super-app with 68M+ users, to get into where crypto really gets used vs where it just gets speculated on.

Live on X Spaces this Thursday, 2:30 PM UTC.

Set your reminder below. ⬇️

https://twitter.com/i/spaces/1RKjppqLgVBJw

Don't miss.

In 1997 the web reserved a status code called 402 payment required. then it sat untouched for nearly three decades.

Then the readers stopped being human. 100M+ machine payments have now settled on base, no accounts, no humans, just a request and a price.

The scarce asset in this economy isn't models or compute, it's settlement.

Read the full piece at Cluster Blogs.

Is crypto dead, or are we just getting started?

At Cluster Roundtable, Dave from Api3 weighs in on why the technology is more useful today than ever and where autonomous agents fit into the picture.

Cluster's private inference layer, explained.

More than 500 open source models available through a single API. You get access to everything from one endpoint and pick what fits the task, without creating separate accounts or switching between providers.

When you route sensitive workloads through private inference, your prompt goes in and the response comes out: nothing in between is visible. The process involves hardware-sealed enclaves and signed attestation, which gives you verifiable proof that your data was never exposed during inference.

x402 handles payments at the inference level. Your application pays for exactly what it uses, settled onchain per request, which removes the need for monthly seats or fixed contracts.

Compute runs inside the same stack. It's what powers the model inside the enclave, not a separate vendor, not a separate integration.

This is what it means to have infrastructure that works as one system rather than four services stitched together.

hub.clusterprotocol.ai

Cluster x @datalinexyz

A price with no confidence score is just a guess. An agent trading on raw, fragmented data is trading blind.

Dataline turns one natural-language query into a clean, confidence-scored read across spot, perps, prediction markets, and onchain. We're bringing it to the agents built on Cluster.

We're wiring Dataline's API into CodeXero so clean market data is one call away from anything you build.

Read on how it fits together at Cluster Blogs.

We're excited to be building on Venice, the leading private AI platform, to solve the harder version of the problem: privacy you can prove.

Venice is the only AI platform of its scale built private by default. Prompts are never stored, conversations never leave your device, and no third party (not Venice, not the infrastructure, not anyone) can see what you ask. Inference runs inside sealed hardware enclaves, and every response ships with cryptographic attestation that it stayed private.

Cluster is the orchestration and settlement layer on top: routing sensitive workloads to Venice's verified private compute and letting autonomous agents pay per call over x402 on Base. No account, no API keys, no data trail.

Together it forms the shape the agent economy has been missing: inference that is verifiably private end to end, and payable by machines, settled onchain against proof instead of trust.

Private inference without the trust assumption.

If Web3 is the future then why most people still can't figure out how to use it.

We're bringing on @uxlinkofficial, one of the largest social platforms in Web3, to get into the real reason normal people still aren't here.

With Rolland, live on X Spaces.

🗓 July 23, 2:30 PM UTC.
🔗 Space Link ⬇️

https://twitter.com/i/spaces/1qJVmmBNgPDGB

Don't miss.

Before AWS, every company ran its own servers. The integration cost was identical everywhere, the advantage was zero, and then one API made it a utility. AI infrastructure in 2026 is the same chaos, one layer up. Hundreds of models, dozens of APIs, no common settlement.

The Amazon moment for AI is not just a better API. It is the collapse of interface, market, and settlement into one layer, where models, data, and compute are purchasable in one request, settled onchain, with no accounts required.

We wrote about why this moment is overdue and what open settlement unlocks that a billing page never could.

Read full article at Cluster Blogs.