Elytro
Self-custodial smart account wallet for AI agents
Performances
RD Popularity Index
RD Growth Index
Comparison
Follow Updates
Follow Lists
People
About Elytro
Elytro is an ERC-4337 smart contract wallet built for AI agents and humans sharing the same on-chain account. Its core business: users or agents interact via a CLI (or legacy extension), while policies—spending limits, 2FA, social recovery, allowlists—are enforced by the smart contract itself, not by any server.
Core differentiator: on-chain policy enforcement. Unlike EOA wallets that assume a human confirms every transaction, Elytro lets agents execute within preset rules (e.g., daily caps, allowlisted addresses, required approvals). The contract snapshots balances before each call and tracks gross outflow, preventing later inflows from masking earlier spending. All code is open-source and audited (2024/2025).
Market pain: traditional wallets can't safely delegate to AI agents; users fear uncontrolled spending or key loss. Elytro solves this by making limits and recovery programmable on-chain, with no seed phrase needed—social recovery via trusted contacts.
Recent 6 months (Feb–Aug 2026): expanded L2 support (Optimism, Arbitrum, Base), daily gas sponsorship for first 10 transactions, stablecoin gas payment, and CLI as primary interface. Negative: mobile app still unavailable; the Chrome extension was deprecated to legacy status, frustrating some existing users.
Key milestones in Elytro's development as an Ethereum smart contract account wallet:
In March 2026, the Elytro CLI repository was created on GitHub, marking the official launch of the project. The CLI supports multi-account management, transaction simulation, cross-chain operations (Ethereum, Optimism, Polygon, Arbitrum, Base, etc.), and implements SecurityHook (2FA) with daily spending limits, providing secure smart account management tools for AI agents and users.
Subsequently, elytro-core was released as a from-scratch rebuild emphasizing security and formal verification. The project achieved 59/59 tests passing, 3 fuzz invariants, and 6 machine-checked Lean obligations, demonstrating an extreme pursuit of smart contract security. Its core mechanism, "realized-value enforcement," snapshots account asset balances to ensure every AI agent call is strictly bound by spending caps, preventing malicious operations.
These two nodes represent Elytro's evolution from foundational tools to security hardening, laying a solid foundation for its long-term development in the Ethereum ecosystem.
Key progress (2026-03-03 to 2026-09-03) centers on security hardening and agent capabilities. On 2026-06-15, the core repo resolved red-team audit findings: added ERC-777 operatorSend to the measured value-mover set, introduced target-scoped open mode (setOpenModeScoped + setOpenAllowedTarget) to bound agent reach, and re-labelled the 1271 typehash blocklist as best-effort defense-in-depth. The same day, the CLI gained a 'dapp' verb enabling agents to call any Ethereum app via open-mode + JIT (executeAsAgentJIT), with eth_call preflight and realized-value caps—proven live on the Cleave testnet with zero per-app allowlist.
Future focus, per official docs and commits: whole-chain open mode remains with documented operational preconditions; JIT reset-assert is sound under honest allowance() views; per-protocol recipient==self pinning is required for approved protocols. The roadmap prioritizes agent-native execution (EIP-4337 + EIP-1271), on-chain policy (spending limits, 2FA, social recovery), and expanding L2 support beyond Optimism/Arbitrum—all while maintaining self-custody and non-custodial security signatures.
Elytro's core team includes Zeng Jiajun, a core contributor. He is a full-stack engineer with 5+ years of experience, previously at ByteDance (frontend lead for Douyin E-commerce messaging) and co-founder of MetaMail. He holds an M.S. in Software Engineering and is active on LinkedIn and Twitter.