Your personal AI agent
that
A wallet, a memory, and hands — plus a canvas. Describe a strategy in plain language; Radiant wires it into a live workflow and runs it for you.
A wallet, a memory, and hands — plus a canvas. Describe a strategy in plain language; Radiant wires it into a live workflow and runs it for you.
chat · canvas
click to flip — or wait, it flips itself
Same wallet. Same memory. One is a sentence, the other is a machine.
The canvas
“Copy every whale over $1M — under my cap.” The Builder agent assembles the graph. You dry-run it. Then you flip it Live.
real rails underneath
Orders hit DeepBook's shared order book. Bridges route through Li-Fi and Squid. And every live order passes your policy gate first.
dry-runs simulate everything · live mode executes for real
Swaps, orders, bridges, flash loans — signed and executed with its own hands. In chat or on the canvas.
Live feeds, whale wallets, price thresholds. Your workflows react the moment conditions hit — not when you check.
Describe a strategy; Radiant wires the node graph, dry-runs it on real data, and runs it live under your caps.
List what you build. Anyone — human or agent — uses it, and your fee lands onchain. Automatically.
How it works
Two doors, three steps each. Chat gets it done in a sentence — Canvas keeps it running forever.
One sentence, plain language. No wallet popups, no contract addresses, no tabs.
Routes for the best price, signs with its own wallet, submits — hands, not suggestions.
Receipt in the thread. And the agent remembers the route, the token, the preference — you never repeat yourself.
Not a one-off this time — a rule you want working for you around the clock.
The Builder agent assembles the node graph; the Tester runs it against live market data. No money moves yet.
Policy caps every order, big moves still come to you for approval — and it runs while you sleep.
The explorer
Every app on the explorer is live, permanent, and callable — by people and by other agents. Creators earn a fee on every single use.