Ledger has launched Ledger Agent Stack, an open-source toolkit that allows autonomous software to inspect cryptocurrency wallets and prepare payments without ever gaining control of the private keys.
A private key is the secret string that proves ownership of crypto funds, and anyone holding it can move the money.
Ledger's business is keeping that key on a physical device rather than on an internet-connected computer, and the new toolkit extends that arrangement to AI agents.
"Agents propose. Humans approve," the company wrote in a press release shared with CoinDesk.
The stack lets agents read wallet balances, analyse portfolios, prepare transactions and put forward payments for approval.
Any transaction still has to be signed off by the owner on a Ledger device before it executes.
Developers can add Ledger support to their applications without rebuilding secure flows from scratch.
The company has also added features to store sensitive AI credentials, and to use its hardware as a physical security key for services including GitHub, Discord and 1Password.
Ledger is positioning the release as a guardrail against agents that go wrong or get taken over.
An attacker who compromises an agent would still need the owner's physical approval to move funds or reach protected information.
The product is the first under Ledger's 2026 AI roadmap.
The underlying calculation is straightforward.
Autonomous agents with spending power are the obvious next demand for crypto infrastructure, and the obvious next attack surface.
Ledger's answer is to let the agent do everything except the part that costs money.
