0x Protocol

The original design attacked fragmentation rather than custody. Makers signed messages; relayers could host and broadcast them; takers submitted chosen orders to Ethereum for atomic settlement. A relayer could charge for discovery without becoming the holder of customer assets.

That architecture made the contract boundary consequential. Efficient off-chain discovery was useful only if applications trusted the settlement code and the approvals it received. Every later version therefore changed not just features, but the route by which authority reached user allowances.

Protocol v2 introduced AssetProxy modules for ERC-20, ERC-721 and bundled assets. Its AssetProxyOwner could add or remove authorized exchange addresses; ordinary changes carried a two-week timelock, while emergency removal could happen immediately.

Protocol v3 changed the token story. The taker paid a protocol fee in ETH or WETH, attributed to the market maker whose order supplied liquidity. Market makers could form staking pools, and ZRX owners could delegate stake to them; rewards combined liquidity contribution with stake. Those mechanics are historical facts, not a promise that today’s API revenue or every modern fill is distributed to ZRX holders.

v4 consolidated functionality as features behind an Exchange Proxy. A Governor controlled migrations and privileged components, and applications could use limit or RFQ orders. RFQ fills omitted the v3 protocol fee, weakening any simple claim that every 0x trade must feed ZRX staking.

The version label now causes confusion. The current site’s ‘Swap API v2’ is not this old protocol v2. It is a later service generation, and its execution path centers on Settler rather than the old AssetProxy exchange.

In June 2020 the builders introduced the 0x Labs name and positioned Matcha as a consumer trading product. Swap API packaged price discovery and calldata for developers. Open contracts remained inspectable, while routing, keys, plans and service terms became an operated product of the ZeroEx companies.

Current API v2 separates approval from execution. Permit2 or AllowanceHolder receives authorization, while transaction.to can be Settler. The Settler repository says a UUPS-upgradeable registry maps feature NFTs to current implementations and can signal pause; clients are told to discover the active address. ZeroEx can therefore ship routing and implementation changes without changing the ZRX token.

ZRX is the quiet object in this moving stack. The verified 2017 token contract created exactly 1 billion units and contains ordinary transfer and allowance functions, but no owner, later mint, freeze, blacklist, fee switch or proxy upgrade.

Its role did change outside the token: early relayer fees and protocol governance were aspirations, v3 made staking concrete, and the DAO later supplied treasury voting. None of that turns the ERC-20 itself into a claim on API fees, Matcha revenue or a redeemable share of treasury assets.

The 2021 DAO launch was deliberately narrow: binding votes governed a community treasury, while core-protocol control was left for a future stage. A later voting-accounting bug could magnify weight by about 1,000 times; the patch postmortem emphasized that this governor touched neither core contracts nor user funds.

An archived FAQ described a Swiss nonprofit 0x Foundation; current service terms instead name ZeroEx Inc. and affiliates. The terms also disclose a ZeroEx-controlled multisig that manages changes to 0x smart contracts, while the ZRX token itself cannot be altered. That is where the original shared-settlement idea has arrived: applications can integrate a company’s routing service and changing execution contracts while the same 2017 token remains in their users’ wallets. Treasury voting never became a single switch for that whole journey.