1 · Why NoirRail exists
Tokenized real-world assets (RWAs) crossed from pilot to production in 2025. Treasuries, money-market funds, invoices, and private credit moved on-chain in real size. The value moved. The privacy did not.
On a public ledger, a tokenized treasury position is a billboard. Anyone can read a fund's holdings, infer its strategy, watch it rebalance, and front-run its flows. For a regulated institution that is not a feature — it is a disqualifier. The market routed around the problem the only way it could without privacy: by retreating into permissioned walled gardens, where the operator sees everything and outsiders see nothing. The privacy is real, but it is bought by abandoning the open, composable settlement that made the technology worth using.
The third path
NoirRail keeps assets on a public, neutral rail — Stellar — and makes confidentiality a protocol property instead of a platform one:
- Hide the numbers cryptographically (note commitments, never plaintext balances).
- Prove the rules in zero knowledge (a Groth16 SNARK per settlement).
- Hand the auditor a key, not the whole ledger (selective disclosure — Phase 2+).
The result keeps the two properties institutions need but rarely get together: confidentiality and auditability.
Why it is buildable today
This is not speculative cryptography. The substrate already shipped:
- Stellar Protocol 22 (CAP-0059) added BLS12-381 host functions to Soroban — the pairing check a Groth16 verifier needs, running natively on-chain.
- Stellar Protocol 25 (CAP-0075) added Poseidon hashing host functions.
- The Stellar Development Foundation prototyped privacy pools on exactly this stack: Circom circuits, Groth16 proofs, verified inside a Soroban contract.
One correction the build surfaced, worth stating plainly because the original design doc had it wrong: Poseidon host functions are Protocol 25 (CAP-0075), not Protocol 23. BLS12-381 is Protocol 22 (CAP-0059). NoirRail's Phase 0 does not depend on the host-function Poseidon — see Chapter 5 for why we run a matched Poseidon in software instead.
NoirRail extends that proven prototype from a fixed-denomination mixer into a value-bearing, recipient-bound, audit-ready settlement layer for named, regulated assets.
The one-line
NoirRail settles real-world value on Stellar the way a private bank settles in a vault: the books cryptographically sealed to outsiders, and openable — on demand and in proof — to the auditor.
What it is, and is not
Is: a confidentiality layer for tokenized RWAs — shield an existing token into a per-asset pool, move and settle it privately, and (later) disclose selectively to auditors. Settlement infrastructure.
Is not: an asset issuer (it wraps what others tokenize), an exchange or AMM (price discovery lives elsewhere), a mixer for anonymity's sake (disclosure is a design goal, not an escape hatch), or a new L1 (it is a contract suite on Stellar).
Try it: read Chapter 2 for the system shape, then run the end-to-end demo in Chapter 11 to watch a shielded value settle on testnet.
If you change one thing: the scope is the product. Every feature beyond private settlement of value-bearing notes was deliberately deferred (see Chapter 13). Adding an order book or a bridge is not "more NoirRail" — it is a different product.