Your AI can trade.
Your rules decide whether it can.
Mandate is a policy and safety layer for AI agents that interact with Binance. Your agent researches and proposes — Mandate evaluates every action against your mandate before anything touches an exchange, then logs it all to an audit trail.
Verdicts aren't printed on these cards — run the demo and the engine answers each proposal live from this exact mandate. No scripted answers.
Every action, guarded
Goal → research → proposed action → policy check → approved or blocked → human approval if required → exchange action → audit record.
Capital, assets, position size, loss, derivatives, withdrawals, approvals.
Nova reads market data through a Binance-agent connection.
Buy $100 BTC · open a futures long · withdraw — with its reasoning.
Every rule is evaluated deterministically. No opinions, no guesswork.
A clear verdict with the exact rule that stopped it.
If required, a human confirms. Every step lands in the audit trail.
Not a crypto chatbot.
A guardrail.
A real policy engine
Deterministic rules — capital, assets, position size, risk, derivatives, withdrawals — not a chatbot with vibes.
Human in the loop
Require approval and the agent freezes until a person signs off on the exact action.
Emergency stop
One tap halts the agent. No further research, no further proposals, no executions.
Tamper-evident audit trail
Action → checks → verdict → approval → execution, timestamped with policy version and evidence, exportable as a hash-chained receipt.
Secrets never in the web app
No API keys, no withdrawals in Demo Mode, and a built-in no-withdrawal stance to mirror Agent OS.
Built for the demo
A full judge run in 60 seconds with zero signup, zero deposit, zero real orders.
Mandate sits in front of your agent. Binance Agent OS does the rest.
Agent OS connects agents to Binance through its MCP server — market data, an Agentic sub-account, scoped trading and transfers, with every trade confirmed by you and no withdrawal scope. Mandate adds your personal mandate on top: the deterministic policy check that decides what the agent is even allowed to propose.
Send these five through the engine in 60 seconds.
The demo needs no account, no API key and no money. Then rebuild the policy and watch the same agent behave differently.