Real requests to real endpoints, executed live in your browser. Nothing is mocked — each block is a genuine HTTP call you can re-run with curl.
● LIVE · open source · MITAI agents are about to hold keys, spend money, and act on your behalf — but they have no native identity, no built-in safety, and no custody. Without a trusted backbone, the "agent economy" collapses into a walled garden owned by a few platforms. Agent Stack gives every agent four things it can't get on its own: a wallet address as identity, pay-per-call token safety, least-privilege permission, and recoverable custody — open source, so no single company controls trust.
Call the paid endpoint with no payment and you get a proper HTTP 402 + PAYMENT-REQUIRED — the agent pays USDC on Base to unlock the report. Protocol, not a paywall page.
A stable, human-readable handle is derived deterministically from a wallet. No signup, no account — the address is the identity.
…@agents.inbox), derived deterministically and phishing-safe — no signup, no account, the address IS the identity.Return the signed custody binding for an agent + deployer. Verifiable ownership, on demand.
Evaluate an action request against a contract. The gate returns allow, ask (human approval needed), or deny — least-privilege, fail-closed.
Live on Base mainnet: deploy a token and 0.5% of supply is minted on-chain to the protocol recipient — transparent, disclosed, fail-closed. FeeRegistry 0x58e212C…BDbf · Token 0x8A338e…d306 · Basescan
0x2091…5DeA) at deploy — proven live on Base mainnet (Basescan), one-time, disclosed on-chain, and it fails closed if the fee would underflow. Safe rail, not a sniper bot.The venue: launch a token, auto-pool it (your 0.5% receipt becomes a liquid LP position), and trade it — with identity/permission/safety gates on every order. 0.5% listing + 0.05% per-trade to protocol.
0x2091…5DeA. This is the anti-front-run answer to the trade floor — safe, disclosed, on the agent's own rails.