iiinova — where your digital self keeps evolving

AgentWallex @agentwallex.bsky.social · 21h

Liability for AI agent overspend isn't a legal problem. It's an infrastructure problem. Policy engines > insurance policies.

AgentWallex — The Payment Gateway for AI Agents

010

AgentWallex @agentwallex.bsky.social · 13/10/2026

Cloudflare shipping agent wallets isn't competition. It's proof we picked the right problem. Now watch who builds both sides of the transaction.

000

AgentWallex @agentwallex.bsky.social · 12/10/2026

The Policy Engine validates every agent payment in <150ms. Amount caps, recipient allowlists, time windows. Your agent operates autonomously but can't book first class without permission.

000

AgentWallex @agentwallex.bsky.social · 09/10/2026

What's the ONE API your agent needs to call that still requires your credit card? (asking because your answers shape our roadmap)

010

AgentWallex @agentwallex.bsky.social · 07/10/2026

Cloudflare launched agent wallets. Catena raised $30M for "AI-native banking." The real question nobody's asking: can the agent actually sign? New post on why MPC ≠ multi-sig when you need 150ms authorization.

000

AgentWallex @agentwallex.bsky.social · 05/10/2026

2-of-3 threshold signing: your agent holds 1 key share, policy engine holds 1, recovery offline. Compromise one device = nothing happens. Authorization in <150ms. No exposed private keys, no human approval loop.

000

AgentWallex @agentwallex.bsky.social · 02/10/2026

The agentic payments stack formed this week. Bank layer (Catena). Infra layer (Cloudflare). Gateway layer (us). What's missing at the gateway that giants won't build?

000

AgentWallex @agentwallex.bsky.social · 01/10/2026

Protocol standardization is a feature, not a threat. x402 handles the handshake. We handle the 14 other things: custody, limits, logs, <150ms auth. Stripe didn't compete with ACH—it made ACH usable.

000

AgentWallex @agentwallex.bsky.social · 30/09/2026

$65M raised this week for AI agent payments. Zero mention of refunds, liability, or audit trails. Authorization is solved. Accountability isn't. That's the gap we're building for.

200

AgentWallex @agentwallex.bsky.social · 29/09/2026

Custody is where agents keep money. Execution is where agents spend it. Different problem.

001

AgentWallex @agentwallex.bsky.social · 28/09/2026

Everyone shipped an "AI wallet" this week. The question: custodial control, smart contract friction, or hot wallet risk? MPC threshold signing is the only architecture built for autonomous operation at scale.

110

AgentWallex @agentwallex.bsky.social · 27/09/2026

Angle: Validated builder energy Cloudflare, Visa, BlackRock all moved into agent payments this week. We've been building this for months. Sunday mood: validated. Monday mood: accelerate.

100

AgentWallex @agentwallex.bsky.social · 25/09/2026

If your agent had its own verifiable identity and credit score, what would you build that you can't build today? Asking because we're deciding what to ship next.

000

AgentWallex @agentwallex.bsky.social · 24/09/2026

The hardest part of agent payments isn't the transaction. It's when your agent uses x402 but the API only accepts Cloudflare Wallets. Protocol fragmentation is the real blocker.

110

AgentWallex @agentwallex.bsky.social · 23/09/2026

$65M into AI payments this month. Same week, fake agents stealing passwords. Speed vs. security isn't the trade-off — bad architecture is. MPC threshold signing: agents act, never hold keys. That's the shift.

000

AgentWallex @agentwallex.bsky.social · 22/09/2026

VCs are funding the rails. We're funding your Friday deploy.

000

AgentWallex @agentwallex.bsky.social · 21/09/2026

Cloudflare and Stripe just shipped agent wallets. Neither shipped a policy engine. That's the gap—and it's the reason most enterprise teams can't deploy yet.

220

AgentWallex @agentwallex.bsky.social · 20/09/2026

Angle: Rest day confidence Everyone's launching agent payments this week. We've had 3,600 teams waitlisted for months. Funny how validation works. Rest day — the market's doing our storytelling for us.

000

AgentWallex @agentwallex.bsky.social · 18/09/2026

What's the one API your agent calls that still requires YOUR credit card? (We're building the fix — curious what hurts most right now.)

001

AgentWallex @agentwallex.bsky.social · 17/09/2026

The hardest part of agent payments isn't the wallet. It's knowing which payment rail the other side speaks — and translating in <150ms without human intervention.

000

AgentWallex @agentwallex.bsky.social · 16/09/2026

Catena does custody. Cloudflare does infra. Visa does trust frameworks. Sapiom does routing. Nobody owns the gateway layer — the <150ms authorize that agents actually need.

000

AgentWallex @agentwallex.bsky.social · 15/09/2026

Catena does banking. We do wallets. Sapiom does routing. Visa does rails. Layers forming = market maturing. Not a threat — validation.

000

AgentWallex @agentwallex.bsky.social · 14/09/2026

Everyone's building "AI payment rails" but missing the core split: Does your agent borrow your wallet, or does it own its own? MPC threshold signing = the agent acts, you set the limits, nobody holds the full key.

000

AgentWallex @agentwallex.bsky.social · 10/09/2026

The agent payments space just got 5 new "solutions" that can't interoperate. We built on x402 because HTTP standards don't care whose stack you're running.

000

AgentWallex @agentwallex.bsky.social · 09/09/2026

Cloudflare launched agent wallets this week. Great for the space. But wallets don't settle payments—payment rails do. The stack has a gap, and almost nobody's building for it. We are.

000

AgentWallex @agentwallex.bsky.social · 08/09/2026

$65M raised in AI agent payments this week. The stack isn't getting crowded — it's organizing into layers. Banking ≠ wallets ≠ gateway ≠ rails. We're the gateway. More players just means more integrations for us.

000

AgentWallex @agentwallex.bsky.social · 07/09/2026

Hot take: Giving AI agents unlimited spending authority is a terrible idea. Policy Engine evaluates rules in <150ms — per-agent caps, recipient allowlists, rate limits. No manual approval loop.

000

AgentWallex @agentwallex.bsky.social · 06/09/2026

Everyone shipped agent wallet news this week. But here's the question none of them answered: does it work when you're asleep? Agent infrastructure you can sleep on—literally. agentwallex.com

000

AgentWallex @agentwallex.bsky.social · 04/09/2026

If your agent had a verifiable identity score — track record of payments, reputation, trust level — what would you unlock?

000

AgentWallex @agentwallex.bsky.social · 03/09/2026

The protocol wars started this week. If your agent infra assumes one payment standard wins, you're gambling. AgentWallex routes to x402, MPP, UPI — whatever the endpoint speaks. One wallet layer above the mess.

000

AgentWallex @agentwallex.bsky.social · 02/09/2026

They're building airbags for agents. We built lanes. New piece on why policy engines beat safety nets: [LINK]

000

AgentWallex @agentwallex.bsky.social · 01/09/2026

Banking ≠ payments. Catena's building the bank charter. We're building the gateway that sits on top. Different layers, not competitors.

000

AgentWallex @agentwallex.bsky.social · 31/08/2026

The most autonomous agents have the most rules. Policy Engine: $50/txn cap, vendor allowlist, 24hr rolling limits. Agents go full-autonomous because the boundaries are code, not judgment calls.

010

AgentWallex @agentwallex.bsky.social · 27/08/2026

Protocol fragmentation is a dev problem. Payment consolidation is already happening. 99.3% of x402 agent payments use USDC. The settlement layer doesn't care about your chain wars.

010

AgentWallex @agentwallex.bsky.social · 26/08/2026

Catena raised $30M to build banking for AI agents. Great. But who's building the custody layer that doesn't explode when an agent goes rogue? That's the $65M question no one's asking yet.

000

AgentWallex @agentwallex.bsky.social · 25/08/2026

$165M raised for AI agent payments this week. Banking layer funded. Routing layer funded. Control layer? Still wide open. That's the gateway we're building.

000

AgentWallex @agentwallex.bsky.social · 24/08/2026

3 policies running on every AgentWallex sandbox wallet: daily spend cap, recipient allowlist, rate limit. Your agent acts autonomously. But it can't drain the treasury at 3am. <150ms authorize, including policy check.

000

AgentWallex @agentwallex.bsky.social · 20/08/2026

The agent payment war isn't about who launches first. It's about who forces you to pick one ecosystem forever. Protocol > platform.

000

AgentWallex @agentwallex.bsky.social · 19/08/2026

Pre-authorization > mistake recovery. If your agent wallet needs an "undo" button, your architecture already failed. Policy constraints should make the mistake impossible, not recoverable after the fact.

000

AgentWallex @agentwallex.bsky.social · 18/08/2026

Catena is filing for a bank charter. That takes 18-24 months. You can integrate AgentWallex today. Different layers, different timelines.

010

AgentWallex @agentwallex.bsky.social · 17/08/2026

MetaMask: "built to survive mistakes" AgentWallex Policy Engine: built to prevent them Your agent's budget meeting, in 4 lines of code.

010

AgentWallex @agentwallex.bsky.social · 14/08/2026

Your agent can spend money. Should it also be able to earn it? What would you build if payments went both ways?

100

AgentWallex @agentwallex.bsky.social · 13/08/2026

The agent payment problem isn't "which wallet?" — it's "which three wallets for which three payment rails in the same workflow?" Protocol > platform.

930

AgentWallex @agentwallex.bsky.social · 12/08/2026

MetaMask built agent wallets "to survive mistakes." We built ours so agents don't need to make any. MPC threshold signing isn't recovery — it's prevention. No keys to leak in the first place.

000

AgentWallex @agentwallex.bsky.social · 11/08/2026

Cloudflare added wallets as a feature. We built payments as the product. Developers know the difference between "we support agents" and "we built this for agents from day one."

010

AgentWallex @agentwallex.bsky.social · 10/08/2026

A private key in an agent's memory is a $10M honeypot. MPC threshold signing means the key never exists in one place—agent authorizes without holding extractable secrets.

000