What to fork, and what to write

1. The headline finding

That is worth stating plainly because it reframes the question. "Fork the platform" is not available. What is available is forking the hard parts individually — and some of those parts are genuinely worth taking rather than writing.

Who already runs this model

The commercial operators confirm the model works and that none of it is open:

The one nominally adjacent open-source project — a self-hosted collectibles vault for comics and figures — is a personal inventory tracker with valuation lookups. It has no order book, no ledger, no custody model and no title transfer. It solves a different problem that happens to share a noun.

2. The pieces that are forkable

Decompose the product and each layer has mature open-source options. Licences below were read from each repository, not recalled — the difference between MIT and GPL decides whether a fork is viable for a commercial product at all.

Languages and datastores were read from each repository's build manifest — build.zig, go.mod, pom.xml, package.json — not recalled. Licences were read from the licence file at repository head. Both can change; re-check before committing to a fork.

3. The recommendation

4. What migrating the ledger would involve

Our journal already uses the account and instrument shape these systems expect, which is why this is tractable:

// ours
{ account: "holder:usr_4f2a17", instrument: "ASSET:itm_c8a1f9", dr: 1 }
{ account: "cash:usr_9b71c4",  instrument: "USD",             dr: 42945000 }

// TigerBeetle: accounts are typed by ledger id, transfers are the entries
account { id, ledger: USD | ASSET, code: holder | cash | revenue, flags }
transfer { debit_account_id, credit_account_id, amount, ledger, code }

// The asset leg and the cash leg become linked transfers in one atomic
// batch — the same all-or-nothing property our tx() wrapper provides, but
// enforced by the database instead of by a JSON snapshot and a try/catch.

Two things would have to be rethought rather than mapped. The hash chain is ours and would stay ours — these ledgers give immutability and tamper-evidence, but the chained digest that lets an outside party verify history against a head they already hold is an application concern. And non-fungible assets are an unusual fit: a ledger built for currency balances has to model one specific slab as a one-unit account, which works but wants care.