Inference Room launched last week as a launchpad for AI Agents and AI Agent infrastructure. Its first resident, Tack, shipped the same day: a storage and memory layer built for software that runs without human supervision, released straight to general availability with no waitlist in keeping with Inference Room's ship-now ethos.

Across AI and crypto, teams are building AI Agents and the infrastructure those Agents need. ERC-8004 proposes a trust standard, x402 proposes payment rails, and Agent-native APIs are starting to replace tools designed for humans. Most of it is still in the design phase. Inference Room is live.

"Most of what's being built for AI Agents today still assumes a human is sitting next to the Agent, opening accounts and entering credit cards on the Agent's behalf. We built Inference Room because products and tools in this space spend too long in heads and on paper, not in production. Every product we ship goes live the day we announce it, and works for the Agent without a human in the loop. Tack is the first. The next is already in production," said Joaquin Mendes, COO of Inference Room.

Tack is a storage layer that AI Agents can use without a human in the loop. There are no API keys to provision and no accounts to register. Agents pay-per-pin in USDC and get versioned, addressable access to files, state and memory through an Agent-native API. A 5MB pin for one month settles at $0.0010 USDC. Every other cloud storage product requires an account, an API key and a credit card, none of which an Agent can set up alone.

Tack also shipped a second storage track today for private wallet-gated objects: the state an Agent accumulates between runs, readable only by the paying wallet.