Ledger launches Agent Stack to secure AI-driven crypto transactions

Ledger, the Paris‑based hardware wallet maker, unveiled Ledger Agent Stack, an open‑source toolkit that enables AI agents to read crypto balances and draft transactions while keeping private keys isolated on hardware devices. The stack separates the proposal function of an AI agent from the signing action, requiring a human to approve each transaction on a Ledger signer. Read‑only operations such as balance checks run autonomously; any move of funds must be confirmed on the device, protecting against prompt‑injection attacks and key exposure.

The toolkit includes a Device Management Kit, command‑line interfaces for developers and enterprise treasury teams, and OpenPGP encryption that shields agent secrets. Partners like MoonPay and Shisa have already integrated the kit. Ledger frames the launch as a response to research showing that human error contributes to roughly 60 % of security breaches and that a quarter of AI‑agent skills contain vulnerabilities. By anchoring every signing decision in hardware, Ledger aims to contain damage if an agent’s software environment is compromised.