Circuits (Arcis)

A circuit describes the computation the MPC network runs over encrypted inputs. You author it, compile it to a compact arithmetic IR, simulate it locally, and deploy it to the on-chain ComputationRegistry.

The model

A circuit is a function over field elements built from a small instruction set — add, multiply, subtract, constants, and comparisons — assembled into an arithmetic IR. An optimizer folds constants and removes dead nodes; a simulator evaluates it locally so you can test without any nodes.

// Compiled by `confidevm compile`. The body runs entirely inside MPC.
#[confidential]
fn order_notional(price: u64, quantity: u64) -> u64 {
    price * quantity
}

The confidevm CLI

# list bundled example circuits
confidevm list
 
# compile a circuit to the binary IR (+ .abi.json)
confidevm compile order_notional --out order_notional.bin
 
# inspect gate count, inputs, and outputs
confidevm info order_notional.bin
 
# run it locally over sample inputs (comma-separated) — no nodes needed
confidevm simulate order_notional --inputs 1000,7
# → 7000
 
# estimate the on-chain fee + deadline for a registered definition
confidevm estimate-fee --comp-def $COMP_DEF --rpc $RPC --fee-oracle $FEE_ORACLE
 
# deploy the definition to the ComputationRegistry on Base Sepolia
confidevm deploy-circuit order_notional.bin \
  --rpc $RPC --private-key $PK --registry $COMPUTATION_REGISTRY
# → compDefId = 0x…

From circuit to computation

1. compile your circuit to the binary IR.
2. deploy-circuit to get a compDefId.
3. An operator creates an MXE binding a cluster to your compDefId.
4. Clients commission computations against that MXE.

The node fetches the bytecode from the ComputationRegistry and evaluates it over the decrypted-in-cluster inputs — so the gate count you see locally is exactly what runs and what you pay for.

Continue to running a node to operate the compute side.