Your intent,
written into rules.
Choose the assets and protocols an agent can use. Set spending limits, risk boundaries, and the moments that need your approval.
A mandate exists before an action does.A policy layer for autonomous agents Live on Robinhood Chain · real funds
Your intent. Their autonomy.
One clear boundary.
An instruction tells an agent what to do. A mandate defines what it is allowed to do. Set the scope, establish the limits, and make every proposed action answer to your intent.
The path of an action
From intent to a decision you can inspect.
Follow the boundary, all the way through.
Choose the assets and protocols an agent can use. Set spending limits, risk boundaries, and the moments that need your approval.
A mandate exists before an action does.A proposed action is checked against the mandate. An asset outside the allowlist, excess exposure, or a breached limit changes the outcome.
Within limits, review required, or blocked.See which rules were applied and why a decision was made. Every decision is recorded against the exact mandate version that produced it, and signed so it can be verified independently.
Traceable intent, from rule to result.The rules are yours
Give an agent a defined space to act.
Keep control over what matters.
Allow specific assets and protocols. Everything outside the mandate stays outside it.
Set per-action and daily caps. Bound concentration, liquidity, slippage, and daily loss.
Require human review above a threshold. Pause a mandate when the circumstances change.
Clarity at the decision point
A result should explain itself. Inspect the proposed action, its mandate, and the rules that shaped the outcome.
Select a scenario to see the boundary in action.
Write your own mandateInside the mandate. Every displayed check passed.
Where intent meets autonomy
A common policy model for the workflows
we are designing MANDATE to support.
Give routine allocations a defined budget, a clear scope, and a point for human review.
Express exposure and execution limits before an agent proposes its next move.
Bring a shared language of permission and decision tracing to agent workflows.
A little more context
MANDATE is a policy layer for autonomous agents. It turns an owner’s intent into explicit rules, checks every proposed action against them on the server, and signs the resulting decision so it can be verified later. It runs on Robinhood Chain.
Sign in with a wallet, write a mandate, and propose a transfer. MANDATE reads the balances and network fees itself, checks the action against your rules, and returns the full rule trace with a signed receipt. This runs on Robinhood Chain mainnet, so an action you sign moves real funds. Start with an amount you are willing to lose, and set your caps before you propose anything.
Not by itself — it holds no key. An allowed action still has to be signed by you, in your own wallet, and the policy is re-checked against fresh state immediately before that. Afterwards MANDATE compares what actually landed on chain against what it authorised, and flags any mismatch.
You set an approval threshold. An action that satisfies every limit but exceeds that threshold waits for a person, and the request expires if nobody decides. An approval binds to that one action under that one mandate version, and it can never overrule a hard limit or an emergency pause.
Yes. Saving changes writes a new immutable version; earlier versions are never edited. Every decision records the exact version it was judged against, so a past result always reads against the rules that produced it — and a new version invalidates approvals still waiting under the old one.
Autonomy starts with intent