Ledger introduced the Ledger Agent Stack, a set of open-source tools that allows artificial intelligence agents to interact with cryptocurrency wallets and blockchain applications. The goal is to automate tasks without taking control of their assets away from the user.
With the new feature, AI agents will be able to check balances, prepare swaps, organize operations and suggest transactions. However, no movement will be executed automatically.
Each operation will depend on the user's explicit approval on a physical Ledger device. This way, even if the AI prepares the transaction, the final decision will remain human.
The launch follows the growth of artificial intelligence in the cryptocurrency market. Companies in the sector are incorporating agents to automate tasks, but Ledger chose a more conservative model.
While some wallets already allow bots to execute operations within rules defined by the user, Ledger limits agents to creating proposals. The transaction signature remains protected by the company's hardware.
According to the company, this approach reduces the risks of AI errors and also makes attacks that attempt to take control of digital wallets more difficult.
"As the attack surface involving AI agents increases, defense strategies lose strength, relying excessively on software firewalls, LLM protections and input filtering," said a Ledger representative. "Even major corporations such as Starbucks and Meta are disabling their AI agents due to privacy concerns and malicious agents."
The company says that threats related to artificial intelligence are growing more quickly than protection mechanisms. According to its data, 26,1% of the skills used by AI agents have at least one security vulnerability.
At the same time, Ledger highlights that AI can also reduce human failures. The company estimates that about 60% of security breaches are caused by mistakes made by users themselves.
For this reason, the company advocates a hybrid model. In this format, artificial intelligence helps prepare operations, while the user remains responsible for the final authorization.
"By anchoring every signing decision in hardware, the Ledger Agent Stack was designed to ensure that, even if an agent’s software environment is compromised, funds cannot be moved and confidential information cannot be accessed without the explicit human confirmation of the Ledger signatory," it added.
In addition to the cryptocurrency market, Ledger intends to expand the use of its devices among developers. Support for the OpenPGP standard allows the hardware to be used to protect API keys and AI agent credentials.
The devices can also function as authentication keys for platforms such as GitHub, npm, 1Password and Discord, offering an additional layer of security.
The Ledger Agent Stack also includes a set of instructions in Markdown to integrate the hardware into the workflows of intelligent agents. MoonPay and Shisa.ai have already implemented this feature in their products.
The launch is part of Ledger's artificial intelligence roadmap for 2026. In the coming months, the company intends to introduce new tools focused on digital identity, agent execution and mechanisms for proving human presence. Founded in 2014, in Paris, the company is also considering carrying out an initial public offering, which could raise its market value to more than US$ 4 billion.
