Project Pine is a research prototype built by the New York Fed’s Innovation Center and the BIS Innovation Hub Swiss Centre, published on 14 May 2025. It asks one question: if bank money and securities move onto tokenized ledgers, can a central bank still do its job? 

The prototype says yes, using smart contracts to set interest rates, buy and sell bonds, and lend against collateral. It is not a plan to automate the central bank, and no central bank has committed to using it.

The Plain-English Version

A central bank steers the economy mainly through three tools. Pine turns each into software.

• Interest on reserves. Commercial banks keep accounts at the central bank, called reserves. The interest paid on them sets a floor under market rates.
• Open market operations. The central bank buys or sells bonds to add or remove cash from the system.
• Collateralized lending. In a squeeze, banks borrow cash from the central bank and pledge assets as security. The central bank discounts those assets by a “haircut” to protect itself.

Pine’s reserves are tokens on a shared ledger, and each tool is a contract that executes instructions the central bank has set.

How the Prototype Is Built

The system has three layers:

1. Settlement layer: Hyperledger Besu, a permissioned Ethereum-compatible blockchain, with blocks every second.
2. Asset layer: ERC-20-based tokens for reserves, deposits, securities and loan claims. Reserve tokens are restricted, so only authorized holders can use them, and they are fungible.
3. Protocol layer: six kinds of smart contracts that carry out the policy tools.

A facility factory sits at the core of the protocol layer. It is not itself a smart contract but a program that creates facility contracts on request. For each facility, the central bank sets the assets exchanged, the borrowing rate, the term, the collateral rules and who may access it. Those parameters can be changed later, and several facilities can run at once. The report is clear on one limit, if a needed parameter was not built into the factory, changing it “is unlikely to be immediate.”

How Each Tool Works

Interest on reserves. Every second the interest contract reads each holder’s balance and accrues interest. Accrual and payment are separate, so the central bank settles when it chooses. The report gives examples: 5% for one holder and 4.5% for another, or 5% only on the first $1 million of a balance. 

Simple positive and negative rates are also supported. Interest accrues on balances, not on unpaid interest, so frequent settlement may be needed.

Open market operations. The report says facilities can be offered through a window or an auction, and the test run used a daily auction at 1 pm. It does not specify bidding rules, pricing or how an exchange of assets for cash settles. Those are open design choices.

Lending against collateral. A bank requests an amount and submits a “collateral pool” contract. The facility checks that the bank has access, that the pool uses the central bank’s pricing, eligibility and haircut contracts, and that the value after haircuts is enough. It creates an exchange contract, and pledging the pool represents acceptance. Prices update every second. Banks can swap collateral in and out if the value is sufficient. When value falls, because of a price drop, a higher haircut or lost eligibility, the contract triggers a collateral call. 

The central bank can set that trigger early, with 3% and 5% given as examples. Liquidation is not automated, if a bank defaults, the central bank manages that with the counterparty.

The Stress Test

Each simulation ran 240 steps (about 10 minutes of real time) with simple automated banks. The report varies five parameters: 

• Event (a routine rate rise or a government bond collapse), 
• Time period (24 hours to 24 months), 
• Liquidity (scarce or ample reserves),
• Number of counterparties (six or sixty),
• Type of credit system. 

The BIS describes ten hypothetical scenarios built from historical data.

The report’s detailed example is a bond collapse with one central bank, six banks, scarce reserves and a 6% target rate:

The lesson is that the early tweaks failed and the new facility worked. The speed came from the factory being able to launch a tool at 5 am, not from automation making the decision.

What Advisers Required

Central bank advisers set the prototype’s requirements: 

• Flexible access controls, 
• Positive and negative rates, 
• Tiered and holder-specific rates, 
• Multiple facilities, 
• Auctions and windows, 
• Collateral baskets and substitution, 
• Changeable haircuts,
• Automatic collateral calls,
• Ability to buy and sell assets. 

The seven central banks that advised were the Reserve Bank of Australia, Bank of Canada, Bank of England, European Central Bank, Bank of Mexico, Swiss National Bank and the Federal Reserve System.

The Risks the Authors Name

• An error in an automated workflow could have “an immediate and wide impact” across the system.
• The price feed could fail or be attacked. The central bank acts as the price source because counterparties trust it.
• Weak privacy could expose which banks use backstop facilities and stigmatize them.
• Tokenization could make reserve demand harder to predict.
• Intermediation could concentrate, and the transition could be long and risky.
• Smart contracts are unlikely to help tokenized and conventional systems work together.

What Pine Does Not Cover

The scope is domestic, single-currency and wholesale. It leaves out financial stability tools, cross-currency operations and signaling tools. It assumes a sound legal basis and does not analyze the law. The simulated banks are simple reflex agents, not a model of a real economy.

The project is limited to research and experimentation and does not reflect Federal Reserve policy.

Pine only applies if wholesale money and securities are tokenized. Two projects are building toward that. Project Agora ran live trials in July with 30 transactions worth about CHF 800,000. 

The European Central Bank launched Pontes on 21 September to settle tokenized assets in central bank money. Both are small, but if they scale, a central bank needs a way to run policy on those ledgers, and Pine is an early answer.

The word “automation” misleads here. Officials still set every parameter, so Pine speeds up and sharpens what central banks already do. The better question is what code adds in a crisis and what it takes away. The annex shows the benefit, a new facility launched at 5 am with no new paperwork. It also shows the limit, because the first two fixes did nothing and judgment chose the third. A system that carries out instructions well still needs humans to choose them well.

Two open points need regulators, not engineers. First, emergency lending in some jurisdictions requires explicit, discretionary approval. Whether a pre-built facility fits those rules is a legal question the report sets aside. Second, the stigma risk. Backstop use is sensitive today, and a ledger that records every draw makes privacy design a policy issue, not a technical detail.

Pine is a credible proof of concept that is open about its limits. It is a contingency plan for a tokenized future. 

Sources: BIS media release, 14 May 2025; BIS Papers, Project Pine; New York Fed, Project Pine page and full report; BIS release on Agora (27 May 2026); ECB press release on Pontes (21 September 2026); Ledger Insights on the Agora live trials (30 July 2026).