Hi — trade-router-mcp is a sharp piece of infra: a non-custodial swap + limit-order engine that AI agents drive directly. That's also the exact point where an agent's bad call becomes an irreversible on-chain swap.
I run a live crypto trading bot (invinoveritas) whose one non-negotiable is a pre-trade governance check before any entry — sized to the account, flags the over-leverage / wrong-size / hallucinated-route cases. I packaged that gate as a single call, /review: returns approve / approve_with_concerns / reject + the specific risk. It's advisory — never blocks the agent, just catches the account-killer before the swap fires.
Could sit in front of the router's execute path, or be exposed as an optional check to the agents using it. One MCP/HTTP call, ~$0.19/call (pennies in USDC via x402), free to try — happy to wire a reference integration myself.
Worth exploring? Even a "not for us, because X" is useful feedback. Nice work either way.
— Federico Blanco, invinoveritas (https://api.babyblueviper.com)
Hi — trade-router-mcp is a sharp piece of infra: a non-custodial swap + limit-order engine that AI agents drive directly. That's also the exact point where an agent's bad call becomes an irreversible on-chain swap.
I run a live crypto trading bot (invinoveritas) whose one non-negotiable is a pre-trade governance check before any entry — sized to the account, flags the over-leverage / wrong-size / hallucinated-route cases. I packaged that gate as a single call,
/review: returnsapprove/approve_with_concerns/reject+ the specific risk. It's advisory — never blocks the agent, just catches the account-killer before the swap fires.Could sit in front of the router's execute path, or be exposed as an optional check to the agents using it. One MCP/HTTP call, ~$0.19/call (pennies in USDC via x402), free to try — happy to wire a reference integration myself.
Worth exploring? Even a "not for us, because X" is useful feedback. Nice work either way.
— Federico Blanco, invinoveritas (https://api.babyblueviper.com)