Ledger launches hardware-secured Agent Stack

Ledger released Ledger Agent Stack on Thursday. The open-source toolkit lets AI agents inspect wallet balances, draft swaps and propose transactions, but prevents agents from executing any action without the user's physical approval on a Ledger device.

Every signing decision must be confirmed on the hardware before a transaction can proceed. The Agent Stack includes open-source Markdown instructions that show developers how to connect Ledger devices to agent workflows. MoonPay and Shisa.ai have added the feature to their products.

The company described the design as conservative. Ledger cited research indicating 26.1% of agent skills contain at least one security vulnerability and noted human error contributes to about 60% of security breaches. Ledger characterized the hybrid AI-human approach as a way to reduce risks from both compromised agents and user mistakes.

Ledger highlighted that its hardware supports the OpenPGP encryption standard, which lets developers store encrypted API keys and AI agent credentials on a Ledger device. Devices can be used like YubiKey systems to log into services such as GitHub, npm, 1Password and Discord.

The release is part of Ledger's 2026 AI Roadmap. The company plans additional tools for agent identity and execution and a “proof of human” attestation to ensure sensitive actions are approved by a person. The company is building the capability for customers to deploy bots but will limit agents to proposing actions rather than executing them.

In a statement, Ledger wrote: “By anchoring every signing decision in hardware, Ledger Agent Stack is designed to ensure that even if an agent's software environment is compromised, funds cannot move and sensitive information cannot be accessed without explicit human confirmation on your Ledger signer.”

Ledger was founded in 2014 and is based in Paris. The firm is exploring an initial public offering that could value it at more than $4 billion.