Control is enforced on-chain, not promised.
ELYX is built so that an agent can only do what its owners allowed, and so that the hard ceiling holds even if ELYX itself misbehaves: it lives in the Safe and the ELYX module on Elysium.
Who owns the capital?
The owners of the Safe. Funds sit in vault Safes on Elysium, and only owner wallets can sign Safe transactions.
Who may use it?
Agents you authorize. Each gets an executor key registered as a delegate of the ELYX spending limit module. It is never a Safe owner and can never sign Safe transactions.
Within what limits?
On-chain limits in the ELYX Safe module per asset, per period and to allow-listed destinations, plus ELYX policies evaluated before every signature.
Who approves exceptions?
Owners, with their wallets. Vault funds need owner signatures over the Safe transaction (EIP-712); agent-wallet swaps need a wallet-signed approval.
What happens before a signature?
Validation, deterministic policy evaluation, then simulation on mainnet compared with the declared intent. Any mismatch blocks.
How is it stopped?
Pause an agent, pause all execution, freeze a vault, or revoke the agent on-chain with an owner-signed Safe transaction.
How is it audited?
Every action is written to an append-only, hash-chained activity log, and every settlement carries its Elysium transaction hash.
Agents operate. Owners keep control. The chain enforces the ceiling.
Layered, so no single layer has to be perfect.
Safe smart accounts
Each vault is a Safe 1.3.0 smart account on Elysium mainnet, owned by the account's owners with their threshold. Agents are never owners, so they can never sign or approve Safe transactions.
Bounded executor keys
Executor keys are derived server-side with HKDF-SHA256 from a server secret and the agent id. They are never stored in the database. On-chain, each is bounded by the ELYX spending limit module: amount per period, token and a destination allow-list, all set by Safe owners only.
Deterministic policy engine
Limits, thresholds, assets, protocols, counterparties, actions, time windows, frequency and slippage are checked with fixed rules. No language model is in the decision path.
Simulation and intent matching
Every transaction is simulated against Elysium (eth_simulateV1) before signing. Contracts called, balance changes and slippage are compared with the declared intent. A mismatch blocks execution.
Owner approvals
Spends from vault funds above a threshold become Safe transactions that owners sign with their wallets (EIP-712, no gas). Swaps from an agent's working wallet and agent change requests use wallet-signed off-chain approvals.
Circuit breakers
Pause one agent. Pause all execution for the organization. Freeze a vault. Revoke an agent on-chain, which removes its spending limits from the vault's module.
Loop protection, rate limits, idempotency
Agents that exceed their action rate or repeat the same intent are paused for human review. Requests are rate limited per credential and organization. Idempotency keys make retries safe.
Append-only audit log
Activity events are hash-chained: each event commits to the previous one, so any edit or deletion breaks the chain. Integrity can be verified from the app.
Sign in with Ethereum
Owners sign in by signing a message with a short-lived, server-issued nonce. No passwords. Sessions are signed, http-only cookies.
Hashed credentials
API keys and agent session tokens are shown once and stored only as SHA-256 hashes. Session tokens are scoped to one agent, expire and can be revoked at any time.
Signed webhooks
Every delivery carries an HMAC-SHA256 signature over a timestamp and the body. Receivers verify in constant time and reject stale timestamps.
No seed phrases. Ever.
ELYX never asks for a seed phrase or private key. Owners sign with their own wallets. Anyone asking for your seed phrase is not ELYX.
What ELYX is, and what it is not.
- ELYX software is not a custodian.
Funds are held in your vault Safes on Elysium. Owners can move them with their wallets without ELYX, through any Safe interface.
- Executor key security depends on the deployment.
Executor keys are derived from the deployment's ELYX_SIGNER_SECRET. Whoever controls that secret can act as your agents, within their on-chain spending limits. Keep it in an HSM or a managed KMS, restrict access, and revoke and recreate agents if it is ever exposed.
- Limits are yours to set.
The on-chain spending limits you approve are the ceiling for each agent. Set them to an amount you are prepared to delegate per period.
- No audit or certification claims.
This page describes how the software is designed. It does not claim third-party audits or certifications for ELYX.