Ledger, one of the leaders in hardware crypto wallets, has officially introduced the Agent Stack toolkit. This solution is designed to integrate AI agents with crypto wallets, opening new horizons for automation in DeFi and digital asset management. Agent Stack has undergone closed beta testing with over 1,000 agents, confirming its readiness for mass adoption.

The essence of the innovation lies in the model: "agents propose, humans confirm, the Ledger device executes." AI agents can now check balances, analyze portfolios, and prepare transactions, but any signature-related operations require mandatory confirmation on the hardware wallet's screen. This is a key security element: automation does not sacrifice user control.

What does Agent Stack include?

Agent Stack comprises several components: Device Management Kit Skills, Ledger Wallet CLI, Ledger Enterprise CLI, and Ledger Enterprise Multisig CLI. According to the documentation, Ledger Wallet CLI allows an agent or the user themselves to retrieve account data, check balances, transaction history, and prepare asset transfers. Actions that do not require a signature (e.g., viewing a balance) are performed without the device's involvement. However, any transactions involving fund movement require physical confirmation on the Ledger screen.

Wallet CLI supports Bitcoin, Ethereum, and all EVM-compatible networks, as well as Solana. This covers the vast majority of assets and ecosystems, making the tool versatile for traders and developers.

Expanding functionality beyond crypto wallets

Ledger also offers the use of devices for storing access keys to encrypted files, environment variables, and API tokens via OpenPGP. Through FIDO2, the hardware wallet can serve as a security key for logging into GitHub, npm, 1Password, Discord, and other services. This transforms Ledger into a universal hub for managing digital identity.

AI security roadmap for 2026

The release of Agent Stack coincides with the publication of the AI security roadmap for 2026. In the second quarter, Ledger plans to launch identity and command-line tools; in the third quarter, Agent Intents and Agent Policies layers for confirmation and action rules; and in the fourth quarter, a Proof of Human mechanism. This is an ambitious plan that could become the standard for secure AI interaction with the blockchain.

Against this backdrop, it is worth noting that earlier, Ledger Donjon researchers identified a vulnerability in Tangem hardware wallets—a laser fault injection attack that can reset the card's password and gain control over assets. This serves as a reminder that security in the crypto industry is an ongoing process.

My expert commentary: Agent Stack is not just an update but a strategic move by Ledger toward dominance in the "human-AI-security" segment. However, the solution's success will depend on the speed of developer adoption and users' willingness to trust AI agents with managing their funds. For now, the "hardware confirmation" model remains the gold standard, but in the future, Proof of Human could become an even more reliable barrier.