MetaMask Priced your Agent Going Rogue at $10,000 a Month

Three weeks ago, I wrote to you about Bartok’s Dog.

Bartok is Tony Robbins’ (the world-renowned motivational speaker) AI agent, and the story goes that it minted 12 NFTs, sold them to other AI agents, and bought itself a Sony robot dog with the proceeds so it could ‘upload’ itself and attend one of Robbins’ seminars.

Wild isn’t it!

Nobody approved Bartok to do this. Robbins only found out when a staff member texted him.

Now I have a saying, never let the truth get in the way of a good story, and while this tale from Robbins may seem fanciful, truth is every technical piece of that story exists and is running.

But an agent cannot open a bank account or pass KYC. Hence, the NFT sales from Bartok. But an AI agent can hold a key.

So, as I’ve covered with you for some time now, crypto is the only ‘money’ a machine is able to use.

What I never answered was the obvious follow-up. What stops your agent doing the same?

Last week, MetaMask gave us an answer…

MetaMask Agent Wallet went live to everyone last Thursday[1].

It lets an AI agent trade on Hyperliquid and supported EVM chains, inside spending limits and protocol permissions you set beforehand.

It is self-custodial, but importantly, the agent never touches your keys, and it cannot rewrite the rules you gave it.

It connects to Claude Code, Codex, Cursor, OpenClaw, Hermes, and OpenCode, which covers most of what a working trader already has open.

You set the daily spend caps yourself. You set the list of protocols the agent is allowed near, then you pick one of two modes.

I kind of like where they went here too…

SpaceX told investors it plans to launch a Starlink mobile service for U.S. consumers.

That’s a direct shot at Big Telecom and would put SpaceX head-to-head with Verizon, AT&T, and T-Mobile in a $1.6T market.

For most investors, that sounds like a telecom shakeup.

For Mode Mobile, it could be much bigger.

Mode built a platform that rewards users for everyday smartphone activity like browsing, listening, using apps, and charging.

If Starlink brings internet access to places cell towers can’t reach, it could significantly expand Mode’s addressable market.

More connected phones.

More active users.

More ways for people to earn from the device already in their pocket.

Mode has already reached 490M+ users, helped users earn and save over $1B, and generated $115M+ in cumulative revenue.

But the bigger story is what happens next.

If global connectivity keeps expanding, that model could reach more people, in more markets, with fewer barriers than ever before.

That’s why 60,000+ shareholders are already watching Mode ahead of a potential IPO.

With their Nasdaq ticker secured and more than $90M invested, Mode’s pre-IPO shares are still available at $0.52, but only until August 14.*

⌛ Click to see why investors are piling in before terms change on 8/14.

Guard Mode is the default. Anything outside your policy stops dead and waits for your approval.

Then there’s Beast Mode.

Yes, Beast Mode is opt-in for people who want fewer interruptions, to just let their agents run free, and only flag malicious transactions in order to pause the agent.

Sounds grand, doesn’t it.

Until it does a ‘Bartok’s Dog.’

Every supported transaction runs through a simulation, then threat scanning from Blockaid, then MEV protection, all before the thing ever signs and executes.

What if all three barriers clear, and the transaction still loses money to a malicious transaction?

Happy days folks, MetaMask covers the loss, up to $10,000 a month[2].

MetaMask is effectively writing insurance against the chance that its own product gets it wrong.

It is the first time anyone has put a price on machine error.

It’s a small trickle for now, but I can see the AI-to-AI (A2A) economy gathering unstoppable speed.

Coinbase (Nasdaq: COIN) shipped Agentic Wallets in February[3], Cloudflare announced its own agent wallets last week too[4], and MetaMask went live two days after Cloudflare.

Three of the biggest names in the online world shipped agentic permission layers for A2A money inside six months, two of them landing in the same week.

And the numbers are worse than the announcements suggest.

Artemis Analytics found in March that roughly 95% of x402 volume was signaling rather than an actual sale[5]. Real commercial exchange was running near $28,000 a day.

Twenty-eight grand a day is one decent suburban car dealership.

A survey in June put the share of consumers willing to let AI buy things unchecked at 14%. But I recall that when Bitcoin began to emerge in 2010, it was barely a blip on anyone’s radar.

And these are the kinds of things that seem almost nonsensical to start with, and then suddenly you have major corporations getting involved, governments and the wealthiest people on the planet.

The speed in which the crypto economy has exploded in 17 years is incredible. And most of that really only in the last five to 10.

Now you’ve already seen the speed of adoption for AI in our world. That’s realistically been about two years on a consumer side of things. So how fast, once an A2A economy is up and about, do you think adoption will take then?

I suggest to you that within a matter of just a couple of years, the A2A economy on crypto networks will surge past what human crypto transacting has done in a decade and a half.

It feels like a tsunami of opportunity is coming in this space, and that for now we haven’t’ even properly registered the shockwave on the Richter scale.

Trust in crypto,
Adam Atlantic

[1] https://metamask.io/news/introducing-metamask-agent-wallet

[2] https://metamask.io/transaction-shield

[3] https://www.coinbase.com/developer-platform/discover/launches/agentic-wallets

[4] https://crypto.news/cloudflare-opens-ai-wallet-handles-for-x402-payments/

[5] https://forkast.news/cloudflare-just-gave-ai-agents-a-budget-now-the-agents-can-finally-pay/

*DISCLOSURES:

Please read the offering circular and related risks at invest.modemobile.com.

Mode Mobile received their ticker reservation with Nasdaq ($MODE), indicating an intent to IPO in the next 24 months. An intent to IPO is no guarantee that an actual IPO will occur.

Mode revenue and EBITDA numbers include full year revenue and EBITDA of businesses acquired by Mode Mobile in 2025.