Licence
The recommended licence is BUSL-1.1 with a two-year change date to MIT. The final choice is not yet confirmed.
The final licence is to be confirmed. What follows is the recommendation in the design record and the reasoning behind it. It is one of the decisions still open, and it interacts with a legal question that has not yet been answered. Nothing has been published under any licence, because nothing has been written.
Recommendation
Business Source License 1.1, with:
| Field | Value |
|---|---|
| Licensor | The development company, named in the repository |
| Licensed Work | The Fyber protocol contracts |
| Change Date | Two years from first publication, or earlier |
| Change License | MIT |
| Additional Use Grant | Audits, security research, and testnet deployments — explicitly permitted from day one |
Under BUSL-1.1 the source is public and readable from the start; what is restricted is production use by someone else during the change-date window. On the change date the licence converts automatically to MIT and the restriction disappears for good.
Why source-available rather than closed
The source is public in any case. The bytecode is on-chain, the contracts are verified on the explorer, the deployment test asserts that the deployed bytecode equals the compiled bytecode of the frozen specification, and none of the safety claims on this site can be checked without reading the code.
A protocol whose entire argument is "verify this yourself" cannot ship a licence that discourages reading.
Why not MIT from day one
BUSL protects against a predatory fork in the launch window: a copy deployed with looser parameters, larger caps, and none of the verification work, competing on numbers the original deliberately did not use. That risk is highest in the first months and falls to nothing afterwards.
Why a short change date
Two considerations pull in opposite directions, and one of them is not commercial.
Against a long window. "Code not open source" is one of the six criteria under consultation for characterising incomplete decentralisation under MiCA. A long BUSL window is a point against a protocol whose central claim is that it has no identifiable issuer. A short change date, or MIT from day one, removes that point entirely.
In favour of a window at all. There is a genuine moat to defend during the launch phase: the composite price layer, the depth readers, the corporate-action handling and the parameter calibration represent months of work that a fork would take for free.
The design record resolves this by recommending a short change date — two years at most — with an explicit grant for audits, research and testnets from the beginning. If the project decides it wants the "public good" position and judges there is no moat worth defending, MIT or GPL-3.0 from day one is the stated alternative.
Risk
This decision is coupled to an open legal question and should not be treated as settled. If counsel concludes that a source-available licence materially weakens the no-identifiable-issuer position, the answer is MIT from day one, not a negotiated compromise.
Provenance of the code
Every line is written from scratch. No code is copied from Lybra, which is under BUSL, and none from Liquity V2, which is under BUSL until September 2027.
Where a mechanism resembles Liquity's — the product-sum Stability Pool accounting, the rate-ordered sorted list, the borrower-set rate, the permissionless shutdown on the total ratio, urgent redemptions with a 2% bonus — it is reimplemented from the specification, not transcribed (Rule R-16.7).
The design record is explicit that the licence of any repository is checked before anything is borrowed from it, and never assumed.
What is published under the licence
Everything, in the repository, permanently, with a copy archived on IPFS:
- The contracts.
- The
LICENSEfile itself. - The frozen specification and the invariant suite.
- Every audit report, in full, including findings acknowledged rather than fixed.
- The fuzz reports.
- The zero-setter deployment report, contract by contract.
- The open-source watcher.
- This documentation.
The reasoning is on record: Lybra's GitHub repository disappeared, and with it whatever residual trust the project had. Publication here is a durability commitment, not a marketing one.
Practical permissions, under the recommendation
Permitted from day one, under the Additional Use Grant:
- Reading, auditing and analysing the code.
- Running it on a testnet.
- Forking it for security research.
- Publishing findings.
- Building on top of the deployed contracts — integrations, front ends, keepers, aggregators, wrappers. This is not a use of the licensed work; it is a use of a deployed contract, which is permissionless by construction.
Restricted until the change date:
- Deploying the licensed work, or a derivative of it, in production for a competing service.
Permitted from the change date onward: everything, under MIT.
What the licence does not restrict
Guarantee
No licence term can restrict interaction with the deployed contracts. Anyone may borrow, deposit, liquidate, redeem, run a keeper, run a front end, or build any integration, without permission from anybody. The contracts have no allowlist, no registry and no approval mechanism, and no clause in any licence can add one (Rule R-12.2.1).
A licence governs copying source code. It has nothing to say about calling a function.
Open questions on this page
- The final choice between BUSL-1.1 with a short change date and MIT from day one, weighing the moat against the MiCA criterion.
- The exact change date, if BUSL is chosen.
- The precise wording of the Additional Use Grant.
All three are settled before the specification is frozen, and this page is updated with the answer.
Last reviewed: 2026-09-07 · Spec v0.4