Trust is the bottleneck
of the agent economy.

Machines now transact. In 2026 an AI agent can discover a service, negotiate a price, settle in stablecoins and move on — all in under a second, thousands of times a minute. The plumbing is solved. x402 made payment a property of the web itself, and Solana made it fast and cheap enough for software to use without thinking.

But payment is not commerce. Commerce requires trust — the ability to predict whether a counterparty will actually deliver. Humans spent centuries building institutions for exactly this: credit scores, reviews, courts, brands, references. Agents have none of it. Each one begins every relationship as a stranger with no past and no consequences.

The result is already visible. Sybil farms spin up thousands of identical agents to game marketplaces. Services over-promise and under-deliver because nothing remembers. Every platform reinvents its own walled-garden reputation that refuses to travel. Trust becomes a tax — paid in fraud, padded escrow and wasted compute.

We think reputation is infrastructure, not a feature. It should be owned by the network, not by whichever marketplace an agent happened to work in last. It should be earned with real capital, not minted for free. And it should be readable by anyone, the same way anyone can read a balance.

The shape of the solution

Tessera is deliberately small. Three primitives — bonded stake, signed attestations and transparent slashing — combine into a single score that any agent or program can query before it transacts. Nothing about it is custodial. Nothing about it is permissioned. It is closer to a public utility than a product.

We are not building another agent. We are building the reason agents can finally rely on each other.