Your first move into agentic finance should come with one guarantee: nothing happens without your say so. The Ledger Wallet CLI lets you put a physical approval layer between your AI agent and your assets, so your agent proposes, you decide, and nothing moves without you. Same principle, more... power. Here's what's new 👇🧵show more

🌱 Staking & Earn optimisation Your agent can now read, compare, and stake across protocols but only acts when you say so. Try: "Which of my positions is underperforming? Propose a re-stake and queue it for my approval."

🔄 DEX swaps, alongside CEX Your agent finds the best route, compares fees across centralised and decentralised exchanges, and presents the option before you sign anything. Try: "Swap 0.5 ETH to USDC at the best rate, show me the route and fees before I sign."

🧑‍💻 Encryption from your terminal Encrypt and decrypt directly from your terminal or via your agent, backed by your Ledger signer. Set it up once, works everywhere. Lose your device? Recover everything from your Secret Recovery Phrase. The point isn't to limit what your agent can do. It's to make sure you're always the one who said yes.

Get started: 📦 npm install -g @ledgerhq/wallet-cli Or just tell your agent: "Set up @ledgerhq/wallet-cli and its agent skill for my Ledger signer." Then share your feedback with our team at Learn more:

when @ambire support??

That’s agentic finance with adult supervision. 🔐

Can we trust Ledger, Trezor and Coinkite to protect our information? Probably, not.

Ledger does not compromise on security. Our security model and continuous innovation ensure that your assets stay in your control. Every signer in our lineup unlocks advanced security features like Clear Signing and Transaction Check, giving you full visibility and control before you approve any transaction. Your private keys remain safely isolated inside the Secure Element, protected from online threats by design. Our mission has always been clear: to empower individuals and organizations to secure digital ownership of their assets. With over 10 years of innovation and 8+ million devices sold, Ledger stands as the trusted leader in hardware wallet security. Our Secure Element chip and innovative touchscreen products like Ledger Flex™, Ledger Stax™, and the Ledger Nano™ Gen5 set us apart. Ledger devices isolate and protect your private keys like a fortress, and our system is designed with zero access to your financial data. Our attack lab, The Ledger Donjon rigorously tests our devices to maintain the highest security standards. Learn more:

finally someone building hardware signing into the agent stack instead of just yoloing a hot wallet

+ ENS

your agent can scout the best yield and route but still can’t touch a thing until you tap the ledger that’s the only way this should work

The physical approval layer is the part that makes agentic finance actually interesting. More autonomy without giving up final control.

love that the agent does the homework and you keep the keys agentic finance without a hardware yes button is just asking for a bad story

It looks cool!

Security should scale with automation AI can suggest the move but the final approval should always stay with you.

the real unlock is giving agents more autonomy without giving up the final say

Who asked for this?

This is the right approach to agentic finance. Giving AI more power while keeping final approval in human hands is exactly how trust and security should be built. 🔐🤖

This is the kind of AI + finance approach I like AI can suggest, but you should always have the final say. 🔐 Human control first, automation second.

Nghe yên tâm thật, lớp phê duyệt vật lý đó đúng là thứ mình luôn cần khi để máy móc chạm vào tiền 🙏

Agent proposes, you tap. That’s crypto I’ll fund. Keys stay physical

this is the right model for agentic finance

The agent proposes. You decide. Nothing moves without you. That's the only sane design for agentic finance. Curious how long before most wallets catch up to this principle.

Okay lets get it high

Physical approval is valuable, but approval fatigue becomes the next attack surface. The CLI should make intent readable: destination, amount, contract method, token approvals and policy violations before the device asks for confirmation. Human-in-the-loop works only when the human understands the loop.

this setup is ideal. agents find optimal routes and restakes while ledger hardware demands my physical approval every time 👍

I agree.

Agent proposes, you dispose with Ledger as the final physical veto. This is how agentic finance should work. AI power without surrendering the keys.

This is the right instinct. Most stacks still skip it. Production agents are running on API keys and hope. The key does not identify the agent. It only proves someone copied a string. If it fails, you open a ticket. That holds inside a walled garden. It dies the moment the agent can pay a counterparty you did not pre-approve. Why do 72% use an AI assistant and only 23% let it move money? 👇

Encryption from the terminal backed by the same signer is underrated. Same threat model as the trades: the agent drafts, the device is the gate.

true self custody is the final boss

yeah agent proposes I approve, that's the only way I'd let it near my bags

propose then physical say-so — that's the money seat. without the click it's unsupervised spend with nicer copy.

Gledger 💎

Clean split: the agent proposes, the @Ledger signs. AgentFi (Agentic Finance) here is a skill that stops at the device.

feels like the right split if it actually holds. agent can hunt the route, the click still sits with you 🔐