Ledger introduces Agent Intent beta with hardware approval

Agents can propose; the owner decides

Ledger introduced Agent Intent on 15 October, presenting a way for software agents to suggest transactions while the owner keeps the signing decision on a Ledger device. An agent is software that can follow a task and use connected tools. In Ledger’s announcement, it can monitor markets and suggest portfolio changes, but the person still authorizes the proposed action. For people considering an agent alongside a hardware wallet, the important change is this handoff: an automated recommendation reaches an approval process instead of becoming permission to spend by itself.

A free beta with a separate roadmap

The company’s product page describes Agent Intent as an approval portal, rather than an AI agent or a service that holds customer funds. The beta is free for compatible Ledger owners. That describes the software’s price, not the total cost of a transaction: network charges and third-party service fees can still apply. Ledger also distinguishes the beta from planned agent policies, which would allow actions within previously approved rules. Those policies are marked as coming soon, so the announcement should not be read as launching unattended spending.

The page places Agent Intent within a broader collection of developer tools. Its promotional artwork illustrates that wider agent toolkit, including a swap example; it is not evidence that every illustrated action is supported by the beta.

What setup and current support involve

The setup guide supports sending funds today; swap requests and Solana support remain in development. It lists Nano Gen5, Flex, Stax, Nano X and Nano S Plus, excluding the original Nano S. A compatible browser, current device software and Ledger Sync are required. Ledger Sync supplies the account-sharing setup, and the owner chooses which accounts the agent may see.

Registration gives the agent an identity distinct from the wallet’s spending key. The owner compares the displayed agent identifier with the one shown on the physical signer before accepting it. This authenticates the agent submitting proposals; it does not approve those proposals. Each subsequent transfer still needs the owner’s hardware approval. The guide explicitly says this beta uses real blockchain transactions, not a practice environment. Once an approved transfer is executed, the beta label does not make it reversible.

Why the device screen still matters

Ledger’s Clear Signing documentation explains the review layer behind that last step. Blockchain requests are encoded instructions, which can be difficult to understand in their raw form. Clear Signing translates supported instructions into recognizable details such as the action, recipient and amount on the signer’s screen. It changes how a transaction is presented, not the underlying transaction itself. Support depends on the relevant integration and transaction information; readable detail is not a blanket promise for every possible request.

That distinction matters when a conversational agent has already described an intended payment. Its explanation and the transaction being signed are two separate things to compare. The useful question at approval is whether the device’s displayed action matches the person’s intention. Clear Signing helps make that comparison possible. It does not judge whether a portfolio recommendation is sensible, or turn an understandable transfer into a profitable one.

The companion command-line tool has a different scope

Ledger also documents a Wallet CLI, a command-line interface through which software can request wallet operations. Its capabilities are broader than the current Agent Intent queue: the documentation covers Bitcoin, Ethereum-compatible networks and Solana, alongside transfers, swaps and selected earning operations. That list belongs to the CLI and must not be substituted for the beta portal’s narrower support list.

The CLI separates observation from execution. After account discovery, balance and transaction-history queries can run without an attached signer, while operations that move funds require device confirmation. A transfer dry run can estimate network fees before signing; it does not send money. This gives an agent a way to prepare a proposal with costs before asking its owner to proceed. Quotes and available third-party services can vary, so a prepared estimate should be understood in the context of the specific transaction rather than as a permanent price for using an agent.