Natural-language research and testnet swap preparation — every action gated by the Route Guardian and your Policy Console. The AI never sees your keys, never signs silently, and a BLOCKED verdict can never be overridden.
TESTNET MODE · VALUELESS TOKENS ONLY
AI AGENT STATUS
● ACTIVE
Wallet: not connected — visit the Wallet page
SUGGESTION — An idea from the agent. Nothing prepared, nothing signed. You can ignore it.
PREPARED — A real transaction is built and checked (guardian + policy passed), but NOT signed. It only executes after your explicit wallet approval.
COMPLETED — Signed by you and confirmed on-chain. Carries a real transaction hash + explorer link.
NATURAL-LANGUAGE COMMAND
Try: “show my testnet portfolio” · “prepare a swap of 2 USDC to wNEAR” · “is this contract suspicious 0x…” · “find the safest swap route” · “explain my transaction in simple language”.
Automation rules
Rules only create SUGGESTIONS — they can never execute transactions. No silent auto-execution exists anywhere in the agent.
No rules yet.
Agent activity log
No agent activity yet. Ask something above.
The log records plans, guardian verdicts, policy results, and tx hashes — never secrets.
Security & permissions
What the agent can never do: see your keys or seed phrase · sign or execute without your explicit wallet approval · override a BLOCKED verdict · change its own policies or risk thresholds · invent quotes, balances, or tx status.