Here’s a fresh, fully rewritten version with a clean and slightly more compact tone:

AI agents can access wallet balances and assess crypto portfolios, but any critical action must be physically approved on a Ledger device before it can proceed.

Ledger has introduced Ledger Agent Stack, an open-source toolkit that brings its hardware-based security approach into the fast-growing AI agent ecosystem. The system allows autonomous software to interact with crypto wallets without ever taking control of private keys.

With this framework, AI agents can track balances, evaluate assets, prepare transactions, and recommend payments. However, execution is locked behind user confirmation on a Ledger hardware wallet, ensuring that control remains with the owner at all times.

The launch is the first step in Ledger’s 2026 AI roadmap, reflecting its view that human validation will remain essential as AI agents take on more advanced financial responsibilities.

Ledger summed up the model as: agents suggest, users decide. According to chief human agency officer Ian Rogers, this approval-first approach has already secured billions in digital assets and now enables AI agents to operate within the same trusted system.

The toolkit also helps developers connect AI agents to both individual and institutional wallets, while keeping transaction authorization tied to Ledger hardware. It includes built-in tools that simplify adding Ledger support to AI-powered applications.

Beyond crypto, Ledger is extending its security model to protect sensitive AI credentials. Its devices can securely store keys and act as physical authentication tools for services like GitHub, Discord, and 1Password.

Ledger says the goal is to minimize the risk of rogue or compromised AI agents. Even if an agent is breached, any attempt to move funds or access secure data would still require the user’s physical approval on a Ledger device.

If you want, I can make it shorter, more technical, or more punchy for a quick news piece.