Foundation Devices, the company behind the Passport Bitcoin hardware wallet, has raised $6.4 million to push into a space that barely existed a year ago: authorization infrastructure for AI agents.

The idea is straightforward, even if the execution isn’t. As AI agents move from novelty demos to tools that can actually spend money, book services, and interact with APIs on your behalf, someone needs to build the guardrails. Foundation thinks the company that spent years figuring out how to secure Bitcoin private keys is the right one for the job.

From hardware wallets to AI permissions

Foundation Devices was founded in 2020 by Zach Herbert in Boston. The company built its reputation on the Passport, a Bitcoin-only hardware wallet with an open-source design philosophy and a heavy emphasis on minimizing the trust users place in third parties. It also offers the Envoy companion app, which handles transaction management and wallet configuration.

The core competency here is secure key management. Storing private keys, signing transactions, and making sure no intermediary can tamper with the process. That’s the bread and butter.

Now the company is taking that same architectural thinking and applying it to a different problem: what happens when an AI agent needs permission to act on your behalf? Think of it as a policy layer. Instead of signing a Bitcoin transaction, you’re defining and enforcing rules about what an autonomous agent can and cannot do with your digital assets, your accounts, your money.

It’s a subtle but important reframing. Hardware wallets protect keys. What Foundation wants to build next protects decisions.

Why AI agents need authorization infrastructure

Here’s the thing. The current wave of AI agent development has a glaring gap. Most agent frameworks focus on capability, making the agent smarter, more autonomous, better at chaining tasks together. Very few focus on constraint, defining exactly what the agent is allowed to do and giving the human a kill switch that actually works.

In English: we’re building increasingly powerful digital assistants and handing them the keys to our financial lives without a proper lock on the door.

This isn’t a hypothetical problem. AI agents are already being integrated into crypto wallet functionalities, executing trades, managing portfolios, and interacting with DeFi protocols. The question of “who authorized this agent to move $10K from my wallet into a liquidity pool at 3 a.m.” is becoming less theoretical by the week.

Foundation’s bet is that user-controlled authorization policies, essentially programmable permissions that the user defines and that the agent cannot override, will become critical infrastructure. Not just for crypto, but for any context where an AI agent touches sensitive data or financial instruments.

The crypto angle is the natural starting point. Digital assets already operate in a key-based security model. Wallets already use cryptographic signatures to authorize transactions. Extending that model to cover AI agent actions is an architectural leap, but not a conceptual one.

The competitive landscape and what investors are watching

Foundation isn’t the only company noticing this gap. The intersection of AI and crypto wallets is becoming one of the more crowded corners of the startup landscape, with multiple teams building agent-to-wallet bridges, autonomous trading bots, and on-chain AI coordination layers.

But most of those projects are building the agents themselves, or the on-chain rails the agents use. Foundation is positioning itself on the authorization layer, the infrastructure that sits between the user’s intent and the agent’s execution. It’s a different part of the stack, and arguably one with fewer competitors.

The $6.4 million raise is modest by venture standards. But for a company that has historically operated with a lean, Bitcoin-purist ethos, it represents a meaningful expansion of scope. Foundation is moving from a single-product hardware company into a protocol-level infrastructure play.

For investors in the broader crypto-AI convergence, this is a signal worth tracking. The market has so far rewarded flashy agent demos and token launches. The less glamorous work of building permission systems, policy engines, and user-controlled guardrails hasn’t attracted the same attention. That could change quickly if a high-profile agent exploit, say, an autonomous bot draining a wallet because its permissions were poorly configured, forces the industry to take authorization seriously.

The risk, of course, is timing. AI agent adoption in crypto is still early. If autonomous agents don’t gain meaningful traction in the near term, the market for authorization infrastructure shrinks accordingly. Foundation is essentially building the seatbelts before the cars are on the highway.

The counterargument: seatbelts are easier to install before the crash, not after. And if Foundation’s open-source, privacy-first approach to Bitcoin wallets is any indication, the company is comfortable being early to markets that reward patience over hype.