Why Waku for the RFQ
A central quote server would have been easier. It would also have been a single point of failure, censorship and front-running. What Waku gives us instead.
The bots behind every fill: what they watch, how they quote, why they keep a single numeraire balance, and how they rebalance across chains.
The backbone of iLayer is its network of solver bots. They pick up orders and fill them cross-chain, competing with each other so the user gets the best price. Anyone can run one: the RFQ runs on Waku, so there is no registration and no KYC.
A bot keeps an internal order list built from three streams: orders created on the various OrderHubs, orders filled on the OrderSpokes, and requests arriving on the RFQ channel. From these it derives the orders it can fill:
The second category is the safety net of the network: a solver that wins a quote and does not deliver hands its order, at the same amounts, to whoever is faster.
Each RFQ arrives on the rfq topic. The solver prices it with its own logic and market view, including gas on both chains and, for sponsored orders, the cost of submitting the creation itself. It answers on the user’s private topic. If chosen, its address becomes the order’s filler and it gets an exclusive window to execute.
Solvers maintain only a numeraire token as their balance, ideally a stablecoin like USDC or USDT. Every order is consolidated into it. A multi-token order that wants WETH and WBTC and gives SHIB is executed as:
fillOrder() on the Spoke;No inventory of a hundred tokens, no exposure to what users happen to send.
Volumes are not symmetric: more orders flow into a hot new chain than out of it. Bots rebalance their numeraire across chains to keep the right ratios for the volumes they see. Because the balance is a stablecoin, rebalancing is a plain transfer, not a portfolio problem.
A solver earns the spread between the quote it gives and its own execution cost, plus the gas premium on sponsored orders. It is paid on delivery: the Hub releases inputs only after the Spoke confirms the fill. Competition on the rfq topic keeps spreads honest, and the open-after-deadline rule keeps service honest. Protocol fee on top: 0.03%.
Next: the messaging layer solvers listen on, and why it is not a server.
A central quote server would have been easier. It would also have been a single point of failure, censorship and front-running. What Waku gives us instead.