Coinbase Markets said on Aug. 17, 2026 that it had added Cluster Protocol, or CP, to its asset listing roadmap and published the project’s Base contract address: 0x001AAd84c21A5CD4d696C56d44866e9703c43F77.
The update drew attention to a Base-based project that describes itself as an AI infrastructure layer and orchestration protocol for autonomous workflows. As outlined in the source article, Cluster Protocol combines an OpenAI-compatible API, tokenized data markets, GPU compute and x402 agent micro-payments, with a browser-native prompt-to-dApp product called CodeXero built on top.
An orchestration layer for AI workflows
Cluster’s central pitch is that it can serve as an orchestration layer for autonomous AI workflows. The project aims to bring model inference, data and compute resources onto Base, Ethereum’s layer-2 network, then connect them through a single API and a single settlement layer.
The article argues that today’s AI infrastructure remains vertically fragmented. Developers building AI-native applications often need separate integrations for model providers, data sources, compute vendors, billing systems and payment rails. Those interfaces may not work well together, and they can be difficult to combine with onchain systems. Cluster is presented as an attempt to unify those pieces into a production-ready pipeline rather than leaving them as separate components.
The same piece also makes a distinction between a coherent story and a scalable business loop. It says the idea is logically consistent, but turning that structure into a system that can sustain commercial scale is a different test.
How the four-layer stack works
Inference engine
At the top of the stack is the inference engine. Cluster says it routes requests to more than 500 open-source models through a single OpenAI-compatible API. The article notes that the white paper uses a 500+ figure, while the Hub page sometimes shows 160+, pointing to inconsistency in public descriptions.
The inference layer covers text, image, audio, embeddings and reranking. Requests are allocated automatically based on availability, latency and capacity, and the system can switch when a failure occurs. Billing is handled per token, and developers only need to change one endpoint URL to migrate.
Tokenized data marketplace and AI services
The middle layer covers datasets and AI services. Datasets are stored on IPFS for persistence, while ownership is represented by ERC-721 NFTs on Base. Purchases are split automatically by the PaymentRouter smart contract. Creators receive 85% of revenue, with the protocol and referrers each taking a portion.
According to the article, a creator can upload a dataset, fine-tune a model on the same platform and then deploy that model directly into the inference engine. Each later call can route revenue back to the dataset creator under the contract rules.
The article also raises a structural concern here. Early contributors of data and models could gain an advantage that later participants find difficult to overcome, which may lead to declining marginal returns for new entrants.
GPU compute layer
In the compute layer, developers can bring their own models while Cluster provides hosting and fine-tuning resources. The training output is then connected back to the unified inference API.
Settlement layer
At the base of the stack is settlement on Base. The x402 protocol supports native HTTP micro-payments, which means AI agents can pay per request without accounts, API keys or pre-funded balances. ERC-8004 is used to provide an onchain identity for agents.
The source article says this full loop is already in production, with value movement enforced at the contract level. It compares the design to combining a power plant, a grid, electrical devices and a payment system into one plug-and-play environment.
CodeXero as the user-facing entry point
CodeXero sits at the outer edge of the stack and turns that infrastructure into a browser-native application layer. Users can describe an idea in natural language and generate a complete onchain dApp, with Cluster’s inference, data and compute resources consumed during the process.
The article frames that model in direct commercial terms: lower the barrier to building applications, expand the user base, and convert usage into paid calls against the underlying infrastructure. Whether browser-based “describe and deploy” usage leads to durable retention is left open.
Usage figures disclosed by the project
The article points to operating data released by Cluster Protocol. In July 2026, the project ranked No. 4 on Base’s global active user list, according to the article. Over the past 30 days, the team said it logged 130,000 inference payment requests and $79,000 in settlement volume, placing it among the leading x402 servers on Base.
It also said CodeXero has connected more than 300,000 wallets, deployed more than 25,000 onchain dApps, and processed more than 4 billion tokens through its private AI inference layer.
The article flags an important limitation: all of those figures were self-disclosed by the project, and it says no independent onchain dashboard was cited to cross-check them.
It then contrasts the 30-day settlement total of $79,000 with the much larger “autonomous agent economy” narrative around the project. Even allowing for an early-stage buildout, the article says the paid conversion level remains relatively low. It adds that Cluster entered Base’s application Top 5 in mid-June 2026, which suggests real wallets and developers were using the loop in a live network environment, though the paid scale remained modest.
Funding and team size
Cluster Protocol was founded in Delhi, India in 2023 and has raised a total of $7.75 million, according to the article.
Its latest funding round was announced on April 23, 2026. The round totaled $5 million and was led by DAO5, with participation from Paper Ventures, JPEG Trading and Mapleblock Capital. The article says the money is intended to accelerate the development and distribution of CodeXero.
Public materials cited in the article put the team at roughly 11 people. Co-founder Yatharth Jain has recently appeared as CEO in media coverage and financing releases. The article says he has a computer science background, built the first inference system almost from scratch in the project’s early days, and worked on research papers with scientists tied to Zama, Privasea, Edge AI and people associated with Google.
Another co-founder, Prateek Bhatia, was identified in earlier records as CEO or co-CEO. His public background is described as focusing on blockchain and decentralized technology, though the article says role descriptions have varied over time and detailed division of responsibilities remains unclear.
The core team members listed in the article are Nishant Chinchole as COO, Tauqeer Ahmad in network engineering and DevOps, Hamza Rizvi in backend and cloud engineering, Vishnu as an AI researcher, Anjali Rajput in marketing and community, and Prithvi Roy in UI/UX and design. The article says fuller org charts and detailed resumes are only limitedly disclosed in public channels.
It raises one practical question from an investor’s perspective: how a team of around 11 can support full-stack development and operations across the inference engine, data marketplace, compute layer, settlement layer and the CodeXero application layer at the same time. The article says infrastructure may rely heavily on external GPU and TEE suppliers, while recent funding is more directly tied to the application layer.
CP token structure and unlock schedule
Cluster Protocol’s native token is CP, with a total supply stated at 1 billion. The article says the project’s TGE page shows 1B, while an audit report previously mentioned a 5B supply and per-chain caps. The team attributed that discrepancy to a cross-chain CCIP design, and public-facing pages now use 1B.
The token generation event is already live, with an initial market capitalization of $12 million. The allocation structure in the article is as follows:
- Community airdrop, 15%: 15% of that bucket unlocks at TGE, with the remainder released linearly over 12 months after a one-month cliff.
- Ecosystem and grants, 20%: 5% unlocks at TGE, followed by a three-month cliff and 24-month linear release.
- Team and advisors, 15%: no TGE unlock, then a 12-month cliff followed by 36-month linear vesting.
- Staking rewards, 20%: 10% unlocks at TGE, no cliff, with emissions spread over 48 months.
- Treasury, 15%: no TGE unlock, then a six-month cliff, with release timing determined by the DAO.
- Liquidity, 15%: fully unlocked at TGE and placed into trading pairs.
The airdrop uses a Cluster Reputation Score, or CRS, together with a five-dimension verification framework. The article says more than 250,000 participants were disclosed by the project.
CP is described as the payment asset across the loop, used for inference requests, dataset purchases, compute calls, staking rewards and governance voting.
The article says the full liquidity unlock at TGE supports market-making arrangements, while the longer cliffs for the team and treasury reduce early sell pressure. At the same time, the later release schedule remains something market participants will need to monitor.
What the Coinbase roadmap signal does and does not say
The article closes by saying Coinbase’s decision to place CP on its roadmap is an important market signal, but not a final confirmation of listing. In the near term, it says attention is likely to center on listing expectations and token unlock schedules. Over a longer horizon, the key question is whether the combination of AI inference and blockchain can produce a paid loop that grows beyond its current scale.
As framed in the piece, Cluster’s narrative is internally coherent and its architecture attempts to join multiple parts of the AI value chain into one system on Base. The gap that remains is between that design, the project’s disclosed usage scale and the size of the team executing it. The article also points to the disconnect between more than 300,000 wallet connections and $79,000 in settlement volume over 30 days as a metric worth following.
