We Built the Infrastructure That AI Billing Was Missing
Every major infrastructure shift in software history has produced the same pattern: a new primitive becomes economically significant, and the tooling to safely transact around it lags years behind. We saw this with cloud compute, with usage-based SaaS, and with payments. We are watching it happen again with AI — and this time the stakes are higher, because the cost surface is more complex and the trust requirements are more demanding.
WeaveOS exists to close that gap. We are building the billing and settlement layer that AI-native businesses need to operate with confidence — where every workflow is priced before it runs, every outcome is verified before a dollar moves, and every payment settles atomically across all parties simultaneously.
The Problem We Kept Seeing
We spent years watching AI product teams navigate the same set of compounding problems. Not as outside observers — as engineers and operators who built and shipped AI products and felt these problems from the inside.
The first problem is cost opacity. AI workflows are not fixed-cost operations. A single agent task can involve dozens of model calls, tool invocations, retrieval passes, and third-party API lookups — each with its own price. The total cost of that task is not known at the moment a customer is charged. It is discovered later, in a billing statement, after the margin damage is done.
The second problem is settlement friction. When an AI workflow produces a result, multiple parties have a legitimate claim on that revenue: the model provider, the tool APIs, the platform, the operator. Today, these payments are settled through a chain of manual invoices, batched reconciliation, and delayed transfers. The money moves slowly, the accounting is error-prone, and nobody has a single source of truth.
The third problem is outcome trust. In a world where agents produce valuable deliverables — contracts, analyses, decisions, code — there is no standard way to prove what was produced, when it was produced, or whether it met the criteria the customer paid for. Disputes are handled by emailing support. Evidence is anecdotal. Resolutions are inconsistent.
These are not edge cases. They are the default operating conditions for every AI business that has moved past the prototype stage.
What We Built and Why
WeaveOS is a pricing, escrow, and settlement protocol for AI workflows. The core design decision was to treat every workflow as a financial transaction from the moment it begins — not from the moment it completes.
When a customer initiates a workflow, WeaveOS generates a binding quote: a precise price, a set of success criteria, and a locked escrow. The customer knows what they will pay. The operator knows what they will earn. The funds are held on-chain before a single model call is made.
When the workflow completes, a verifier evaluates the outcome against the success criteria defined in the quote. The evaluation is cryptographically attested — signed by a trusted execution environment, with the outcome, trace, and proof stored immutably on Walrus. No one can retroactively alter what was produced or what criteria were applied.
Settlement is then triggered automatically. A single atomic transaction on Sui moves funds simultaneously to every recipient in the payment graph — model provider, tool APIs, operator, platform — with no intermediate steps, no manual reconciliation, and no possibility of partial payment. Either every party is paid in full, or no one is paid and the customer is refunded. That guarantee is enforced at the protocol level.
Why On-Chain Settlement
The decision to build on Sui was not arbitrary. Multi-party atomic settlement — the ability to disburse to many recipients in a single all-or-nothing transaction — is a hard problem in traditional payment systems. It requires trust in an intermediary, or a complex coordination protocol, or both.
On Sui, a single Programmable Transaction Block can transfer to any number of recipients atomically. The settlement logic is a Move smart contract: open, auditable, and unable to be changed by any single party after deployment. The escrow is held in USDC, not a proprietary token. And every settled workflow produces a permanent on-chain record — a proof of payment that any party can verify independently.
This is not about decentralisation for its own sake. It is about using the right tool for the job. When you need to move money to six parties simultaneously, with no manual intervention, and produce a tamper-proof receipt, a blockchain is not an exotic choice. It is the practical one.
Verifiable Outcomes as Infrastructure
The settlement layer is only useful if the outcome verification layer is trustworthy. We spent significant design effort on this, because it is the part most infrastructure providers skip.
WeaveOS ships a success criteria DSL — a structured, machine-readable specification of what a workflow must produce to be considered successful. Operators define criteria at quote creation time: exact string matches, regular expressions, JSON schema validation, numeric thresholds, or boolean compositions of the above. These criteria are CBOR-encoded, hashed, and committed to the quote on-chain before execution begins. They cannot be changed after the fact.
At settlement time, the verifier evaluates the actual outcome against those committed criteria inside a trusted execution environment. The attestation payload — including the workflow ID, outcome blob ID, trace, success verdict, and cost reconciliation — is signed by the enclave and verified by the Move contract before any funds move.
In production, the enclave is an AWS Nitro instance with PCR measurements registered on-chain. The verification chain runs from the AWS root certificate, through the enclave attestation document, to the Move contract — with no trusted intermediary in the path. For the hackathon and early integrations, the same pipeline runs with an ed25519 dev key. The cryptographic structure is identical; only the trust anchor differs.
The Economics We Are Enabling
We are not building a payments product. We are building the economic infrastructure that makes an entirely new category of AI business model viable: outcome-based pricing.
Today, AI products are priced on access — per seat, per API call, per month. This works when the product is a tool. It breaks down when the product is a result. If an agent produces a qualified sales meeting, a completed due diligence report, a working code migration — the economic logic of charging for that outcome is completely different from charging for the tokens consumed.
Success fee pricing — charging only when success criteria are met — is the natural pricing model for outcome-based AI. But it requires infrastructure that can hold payment in escrow, evaluate outcomes objectively, and release or refund atomically. That infrastructure did not exist. WeaveOS is that infrastructure.
Alongside success fees, we support fixed pricing for predictable workflows, capped pricing with markup for operators who want cost certainty, and hybrid models that combine a base fee with outcome-contingent components. The right model depends on the workflow. The platform supports all of them.
Where We Are Going
The first version of WeaveOS is settlement infrastructure: quotes, escrow, outcome verification, and atomic multi-party payment. That is the foundation — and it is running on Sui testnet today.
The layer above it is pricing intelligence: real-time cost tracking as workflows execute, margin guardrails that intervene before a task becomes unprofitable, customer-level profitability analytics, and benchmarking data drawn from the aggregate settlement graph. We are building this now.
The layer above that is the agentic economy's financial primitive: a settlement protocol that any AI agent can invoke to transact with any other party — human, business, or machine — with cryptographic guarantees on both sides of the exchange.
We are early. The category is early. But the problems are real, they are growing with every AI product that ships, and the teams building those products deserve infrastructure that is built to the same standard of correctness they hold their own software to.
That is what we are building.
Get in Touch
If you are building AI products and recognise these problems, we would like to talk to you. WeaveOS is in early access on Sui testnet. Design partners get direct access to the team, input into the roadmap, and settlement infrastructure that is production-ready from day one.
Reach out via the dashboard, through our developer docs, or directly at team@weaveos.xyz. We read everything.