SAFETY: decode an UNSIGNED EVM transaction and flag drain patterns (unlimited/large approval, approve-all, token & NFT transferFrom, proxy upgrade, on-chain permit, approvals hidden in multicall, opaque calldata) BEFORE signing. Returns {decoded, risk_flags, ok}. Read-only; takes no key.

SAFETY: inspect an EIP-712 typed-data message the agent is about to SIGN and flag permit-phishing (ERC-2612, Uniswap Permit2, DAI-style permits granting an unlimited/large allowance). Catches the off-chain drain a transaction check can't see. Returns {decoded, risk_flags, ok}. Read-only; takes no key.

SAFETY: enforce a policy on a proposed agent action (tool call) BEFORE it runs. policy supports forbid_tools/allow_tools, max_value_wei, allow_recipients. Returns {allowed, violations}. Fail-safe: denies on unreadable input. The 'inspect what the agent DOES' gate, not just who it is.

Create or restore a non-custodial wallet; returns the address and the private key (caller keeps the key).

Read a wallet/address balance from the live chain (read-only).

Sign and post a native-coin transfer with the caller's own key. EVM one-call: omit nonce/gas and they are auto-fetched + broadcast (or supply them to control the tx). Solana uses lamports; Bitcoin uses amount_btc.

Sign and post a call to any contract/app function.

Swap tokens via a DEX aggregator with our built-in fee; non-custodial.

Move value across chains via LI.FI; signs the route tx with the owner's key.