The Agent Commerce Gap

MCP lets agents connect to external systems. A2A lets agents talk to each other. AP2 and ACP let agents pay. These are real, important protocols. But they share a structural gap: none of them solve what happens when an agent needs a capability from a provider it doesn’t already know.

The gap is commercial, not technical

The hard problem isn’t invoking an API at runtime. It’s choosing a provider, verifying their credentials, agreeing on terms, holding payment in escrow, verifying the output, and building a track record that improves future decisions. That’s not a protocol problem. It’s an institutional one.

Protocols let agents connect. What’s missing is the venue that lets them transact — with rules, accountability, and history.

Why this matters now

The agent economy is crossing a threshold. Agents are no longer tools that assist humans — they’re actors that transact, procure, and commit on behalf of organizations. When the counterparty is unknown until execution time, you can’t pre-integrate. You can’t pre-trust. You need a governed venue that embeds trust into the transaction itself.

• Pre-integration breaks when the provider is unknown until runtime
• Bilateral trust doesn’t scale beyond known counterparties
• Protocols can’t adjudicate liability or resolve disputes
• The venue that accumulates transaction history first wins