Crowdloans: a new DeFi primitive for borrowing from your social circle, not a bank

TL;DR. A crowdloan is one loan funded by many people who each choose the borrower and name their own rate, with the borrower picking which offers to take, a smart contract holding the money, and the outcome written to the chain. Nothing in that list is new on its own. Lending circles did the choosing, peer-to-peer sites did the many-to-one, DeFi pools did the contract. What is new is all of it at once, with no bank, no platform grading the borrower, and no collateral. That combination behaves like a primitive: a building block a friend, a DAO or a program can use the same way. This post defines it, says where the word came from, and is honest about what it cannot do.

This article is general education, not financial, legal or tax advice. Unlloo runs crowdloans on Ethereum with no collateral, so a lender can lose what they lend. Figures from the contract were read on 5 November 2026.

What is a crowdloan?

A crowdloan is a single loan to a single borrower, funded by many independent lenders who each decide whether to lend, how much, and at what rate. The borrower posts a request with an amount, the highest rate they will pay and a repayment date. Lenders make offers at or under that rate. The borrower accepts the offers they want, the money moves, and the borrower repays everyone through the same contract. No collateral is locked, and the result, repaid on time or not, is public.

Six properties make it what it is, and if any one is missing, it is something else:

• Many lenders, one borrower. Not a pool that lends to everyone, and not one counterparty.
• Each lender chooses the borrower. Nobody is assigned a loan by an algorithm or a credit grade.
• Each lender names their own rate. The borrower sets only a ceiling.
• The borrower chooses the lenders. They accept some offers and leave others, as long as the accepted total reaches a floor.
• A contract holds the money, from offer to repayment, with no company account in between.
• The outcome is public. Repayment or default is written where the next lender can read it.

On Unlloo's version the floor is 80% of the request, offers run from 0% up to the borrower's cap, and the contract gives lenders a seven-day window after offers close for the borrower to choose. Those are our parameters, not the definition. The definition is the six properties.

Why does the word need reclaiming?

Because for four years "crowdloan" meant something that was not a loan. From 2021, Polkadot and Kusama used the word for the way token holders locked DOT or KSM behind a project's bid for a parachain slot, usually for the project's own tokens in return, and got the same DOT back when the lease ended. There was no borrower, no interest and no repayment, which is to say none of the six properties. Polkadot's own documentation now says that "parachain slot auctions and crowdloans have been deprecated" in favour of its Agile Coretime model, so the word is free, and we think it should mean what it says: a loan from a crowd.

"Crowdfunding" is the nearest relative and the useful contrast. In crowdfunding, many people give money to one project and get a product, a reward or equity. In a crowdloan, many people lend money to one person and get it back with interest, or publicly do not. The word has always been one step away from meaning this.

Is any of this actually new?

The parts are old. The combination is not, and it is worth being precise about which part each ancestor had and which it lacked.

Lending circles had the choosing. A tanda, a susu or a chit fund is a group of people who know each other, paying into a pot and taking turns to receive it, and Besley, Coate and Loury described them in 1993 as "observed worldwide". Every member decides whom to sit in a circle with. What they lack is a record a stranger can read, so the trust never leaves the room.

Peer-to-peer lending had the many-to-one. The Bank of England dates the model to the UK: peer-to-peer lending "first emerged in the United Kingdom and the world with the launch of Zopa in 2005", as its 2016 working paper puts it, with Prosper and LendingClub following in the United States. Many savers funded one borrower's loan. But the platform graded the borrower, set or bounded the rate, matched the money and held it, and most lenders ended up with automated slices of hundreds of loans they never looked at. The lender did not really choose, and the platform could, and sometimes did, fail with the money.

Kiva had the choosing and the many-to-one together. Since its first loan in March 2005, Kiva says 5 million people have borrowed through lenders who pick a specific person. What it lacks is a rate: lenders earn nothing, field partners stand between them and the borrower, and the arrangement is philanthropy with repayment rather than credit.

DeFi pools had the contract. Aave and its kind put the money in code with no company in the middle, and that is the part crowdloans inherit directly. What a pool lacks is any choosing at all: deposits are lent to whoever posts collateral, at one rate for everyone, and the only thing a lender knows about the borrowers is that they are over-collateralised. How DeFi loans work compares the two shapes step by step.

A crowdloan is the first shape with all four: people choose people, many fund one, every lender prices their own risk, and the contract, not a company, holds the money and writes the result down. That is the claim, and it is a claim about structure, not about size.

What makes it a primitive rather than a product?

A primitive is a building block that other things can use without asking, and a crowdloan qualifies because it is five functions on a contract that anyone or anything can call. Create a request. Submit an offer. Accept offers. Repay. Claim. On Unlloo's contract those calls are the whole protocol, and the website is one way of making them, not the only way.

That matters because the caller does not have to be a person at a keyboard. A DAO's multisig can create a request and accept its members' offers. A community treasury can be the first lender on every member's first loan, as a policy. A payroll service that pays contractors in USDC could offer each of them a bridge loan against the next invoice. A lending circle could run its rotation as a sequence of crowdloans, so the record travels with each member. None of those need our permission, and none of them need the others to exist. A product is used; a primitive is composed.

The other half of being a primitive is that its output is useful to things that were not part of it. Every finished crowdloan leaves a public, verifiable fact about a wallet: it borrowed this much from these addresses and repaid it, late, or not at all. Any lender, any protocol, any score can read that fact without trusting us. The anatomy of on-chain reputation is about what such facts add up to.

What does "borrow from your social circle" mean in practice?

It means the first lenders are the people who already know you, and the record they create is what lets strangers lend next. A bank underwrites you from documents. A pool underwrites you from collateral. A crowdloan underwrites you from the people willing to put money behind your name, which is exactly how credit worked before banks and still works in most of the world, with one change: the result is written down somewhere everyone can read it.

The first requests on Unlloo's Ethereum contract show the mechanism in miniature. Five requests were posted between 25 and 30 October 2026. One was funded in full, by four lenders at rates between 1% and 4.2%, all under a 5% cap, and the first offer came from someone who knew the borrower. The four requests from wallets nobody knew, with scores of 120 to 233, got no offers at all. How to request a loan that lenders fund goes through each one, and who uses under-collateralized loans draws the rings: family, friends, employer and community lend first; strangers and treasuries lend once there is a record.

What does a bank do that a crowdloan does not?

Three things, and a crowdloan should not pretend otherwise. A bank underwrites: it has the authority and the data to decide, which a crowd has to build one loan at a time. A bank pools risk: a hundred loans from one balance sheet, with deposit insurance behind it, where a crowd lender carries each loan alone. And a bank recovers: collections, courts, and collateral it can seize. A crowdloan with no collateral has none of that. If the borrower does not pay, the lender's only remedy is the record, and the borrower's only cost is the record.

What a crowdloan has that a bank does not is the mirror image. No gatekeeper, so a freelancer paid in stablecoins, a person in a country with weak banks, or a DAO with a treasury can borrow on the same terms as anyone. No geography: five lenders in five countries and a borrower in a sixth. Minutes to set up, cents to run, and a record that belongs to the borrower rather than to the lender. Is it safe to lend crypto is the lender's side of that trade.

What are the open problems?

Four, and they are the problems of the structure, not of any one implementation.

• No recovery. Without collateral, a default is a loss, full stop. Optional collateral is planned on Unlloo, but a crowdloan with collateral is a different trade and not the one this post is about.
• Identity. A wallet is cheap and a person can have a hundred. The record is only as good as the link between the wallet and the person, which is why the first lenders have to be people who know the borrower, and why reputation scores read more than one wallet signal.
• Liquidity. A pool always has money; a crowd has to show up. Four of the first five requests got nothing, and a primitive that works only when a crowd arrives needs crowds, which is a distribution problem before it is a protocol problem.
• Law. Lending to a person across a border raises questions a pool of collateral never did, and the answers differ by country. The terms say who may use Unlloo; they do not answer the general question.

Where does it go from here?

Toward more things calling the same five functions. The website is the first caller, a DAO treasury is the obvious second, and a payroll or invoicing service is the one we would most like to see. The score does not yet read Unlloo repayments, and when it does, every repaid crowdloan will raise the borrower's standing for the next one, which closes the loop the structure is built for. How Unlloo works today lists what is built and what is not.

Whether "crowdloan" becomes a common noun is not ours to decide. The definition is: many lenders, one borrower, each lender choosing and pricing, the borrower choosing back, a contract holding the money, and a public result. Anything with those six properties is one, whoever builds it.

Sources

1. Polkadot Wiki, "Parachains", read on 5 November 2026. "Parachain slot auctions and crowdloans have been deprecated" with the activation of Agile Coretime.
2. Atz and Bholat, "Peer-to-peer lending and financial innovation in the United Kingdom", Bank of England Staff Working Paper No. 598, 29 April 2016. Peer-to-peer lending first emerging with Zopa in 2005.
3. Kiva, "About", read on 5 November 2026. The first loan in March 2005 and 5 million borrowers since.
4. Besley, Coate and Loury, "The economics of rotating savings and credit associations", American Economic Review, 1993. Lending circles as a worldwide institution.
5. Unlloo's CrowdLoanManager contract on Etherscan, read on 5 November 2026. The five public functions, the 80% acceptance floor, the seven-day selection window, and the first five requests and their offers.
6. Unlloo, "What is a crowd loan?", glossary entry, including the difference from a Polkadot crowdloan.