Private underwriting. Public enforcement.
Public blockchains cannot natively support institutional credit decisioning because the inputs are private and the rails are transparent. CreditWeave separates confidential underwriting from deterministic onchain execution through a live V2 registry, CRE workflow, NAV oracle, lending pool, and digest-based audit references.
Architecture target
From confidential inputs to enforceable onchain credit terms
Confidential Inputs
- Borrower financials
- Credit profile
- KYC / AML context
- Asset and macro data
CRE Runtime
- Aggregate data fetch
- Base underwriting model
- NAV freshness check
- AI qualitative analysis
Policy Gate
- Tighten-only adjustments
- Confidence thresholds
- Hard deny rules
- Reproducible outputs
Onchain Decision
- Approval status
- Max LTV and rate
- Credit limit and expiry
- Reasoning and provenance hashes
Enforcement Layer
- Borrow eligibility
- Collateral checks
- Liquidation path
- Portfolio risk controls
Current implementation
The live stack already runs request storage, CRE underwriting, NAV refresh, digest references, and lending enforcement.
Current limitation
The live CRE workflow currently emits conditional approval or denial only, even though the registry supports richer decision states.
Target direction
The architecture is designed to broaden across products while keeping confidential inputs private and public outputs deterministic.
Problem / Why Now
Public blockchains are transparent. Institutional credit is not.
That mismatch has left onchain credit stuck between overcollateralized lending and narrow product design. If real-world credit is going to move onchain, the market needs an architecture that can keep underwriting private while preserving public, deterministic enforcement.
Privacy Mismatch
Institutional underwriting depends on borrower financials, credit data, and compliance context that cannot be exposed on a public ledger.
Capital Inefficiency
Most onchain credit systems solve for transparency by overcollateralizing instead of underwriting, which limits capital efficiency and product range.
Institutional Friction
Compliance constraints, data handling requirements, and weak risk primitives keep serious credit allocators from moving more activity onchain.
Weak Credit Primitives
Transparent lending rails can enforce collateral, but they rarely support private credit judgment, policy enforcement, and auditability together.
Why now
The market is finally catching up to the architecture this category requires.
Confidential runtime environments, better data availability, and growing RWA demand make it possible to build real underwriting systems instead of public collateral wrappers.
Confidential compute environments are now robust enough to support offchain underwriting flows.
RWA interest is pulling more serious credit use cases toward programmable execution rails.
Data providers, compliance tooling, and audit requirements increasingly demand split private/public architectures.
Why CreditWeave
A credit architecture built for private inputs and public execution.
CreditWeave is not an AI scoring widget layered onto DeFi. It is a full-stack decision system: confidential data ingestion, deterministic underwriting, bounded AI adjustments, and enforceable onchain outputs linked to digest and provenance hashes.
Confidential by Design
Borrower data, API credentials, and qualitative reasoning stay inside the confidential runtime instead of leaking into public infrastructure.
Deterministic at Decision Time
CreditWeave uses a deterministic base layer and policy-bounded adjustments so final terms remain reproducible and constrained.
Auditable and Enforceable
Only minimal terms go onchain, but those terms are hash-linked, attributable, and directly enforceable by lending contracts.
Composable with Capital Rails
The architecture plugs into NAV controls, underwriting registries, lending pools, and portfolio-level risk constraints.
What makes this defensible
- AI acts inside deterministic constraints rather than replacing underwriting policy.
- Only minimal decision outputs are published onchain, keeping private credit data off public rails.
- Digest and provenance references create an auditable bridge between confidential analysis and public enforcement.
Input snapshot hash
0x8f...d3a1
Base decision
Tier 2 / approve
AI adjustment
Tighten LTV by 500 bps
Final terms
65% LTV / 8.5% rate
Reasoning hash
0xc1...5fe9
Onchain action
Signed report submitted
Proof of Build
More than a concept page.
CreditWeave already spans the private decision layer and the public enforcement layer. The core pieces of the V2 architecture are implemented across contracts, confidential runtime logic, APIs, and operator-facing interfaces.
Underwriting Registry V2
Stores decision status, terms, covenants, and digest/provenance references for the active underwriting path.
RWALendingPool
Enforces borrowing, repayment, liquidation, reserves, and underwriting-linked eligibility.
NAVOracle and Risk Controls
Supports NAV freshness, segment-level controls, haircuts, and portfolio constraints.
CRE Underwriting Runtime
Runs confidential data ingestion, deterministic underwriting, bounded AI analysis, and receiver-side verification.
Private API Layer
Provides confidential borrower and asset context plus explanation storage keyed by reasoningHash.
Operational Dashboards
Borrower, investor, and admin interfaces are already wired to the protocol flow.
Built today
A working flow from request to enforceable terms.
The stack already supports underwriting requests, confidential processing, NAV refreshes, signed output submission, and downstream contract enforcement. That matters because the hard part of this category is system integration, not just interface polish.
Borrower submits an underwriting request onchain
CRE fetches private borrower and asset context
Deterministic underwriting, policy gates, and NAV checks compute final terms
Decision and reasoning digests are posted for enforcement
Pilot Strategy
Start where feedback loops are faster, then expand outward.
The architecture is multi-asset, but the current implementation is real-estate-first. The near-term strategy should still be narrow and learnable, so CreditWeave is better served by proving confidential underwriting in smaller business credit flows before moving into larger and slower-moving asset classes.
Invoice Financing
Shorter cycles, smaller tickets, and clearer payment events make this a strong early proving ground for private underwriting.
Typical scope
$5k - $15k pilot tickets
SME Revenue Advance
A natural second wedge once the data layer and policy engine have been validated on smaller business credit flows.
Typical scope
$10k - $25k pilot tickets
Real Estate Expansion
Real estate remains a strong long-term market, but it likely comes after the system has earned trust through faster learning loops.
Typical scope
$50k+ expansion path
Why this wedge
Credibility comes from learning speed, not maximum TAM on slide one.
Lower ticket sizes reduce capital required for early validation.
Faster repayment and default signals create tighter learning loops.
Short-cycle assets make it easier to compare policy outcomes against reality.
Real estate remains compelling, but it is a better expansion market than a first proving ground.
Risk Controls
Built to reduce model risk, data risk, and operator risk.
CreditWeave should not read like “AI decides loans.” The architecture is designed so that private underwriting can still be policy-constrained, reviewable, and enforceable.
Deterministic Base Layer
Binding decisions originate from reproducible underwriting logic instead of free-form model output.
Bounded AI Influence
AI can tighten terms and add qualitative context, but it operates inside explicit policy limits.
Hash-Linked Audit Trail
Reasoning, provenance, and decision traces can be linked back to onchain outcomes without exposing raw data.
Safety Denials and Checks
Stale NAV, asset status failures, compliance flags, and invalid requests can halt underwriting before funds move.
Timed Reviews and Expiry
Credit terms can expire, refresh, and be re-evaluated instead of persisting indefinitely on stale assumptions.
Onchain Enforcement
Final terms are enforced by contracts, not by operator discretion after the fact.
Business Model
Value capture from underwriting infrastructure, not just a front-end product.
The business case is to become the confidential decision layer behind onchain credit. That creates room for fee capture at underwriting, financing, and infrastructure integration points as the system matures.
Underwriting and Origination Fees
Charge for confidential underwriting workflows and the conversion of private data into enforceable credit terms.
Protocol-Level Financing Fees
Capture value from funded credit activity once lending rails and capital pools are active.
Infrastructure Layer Positioning
Offer confidential credit decisioning as reusable infrastructure rather than a single closed lending interface.
Capital and Distribution Partnerships
Expand through originators, allocators, and structured-credit partners that need private underwriting rails.
Roadmap
Implemented baseline, then hardening, then pilot, then expansion.
The next phase is to harden the credit stack, prove it in a disciplined pilot, and then expand into larger credit categories with better evidence.