Mandate Control

An agent is alive, operating under a mandate, making constrained decisions, and leaving behind evidence nobody — including its operator — can rewrite.

mode · operator
Mandate builder · dry-run preview

Adjust the four policy knobs, then dry-run against the canonical France–Sweden fixture. No account, no wallet, no execution.

Minimum edge?Smallest edge (AI fair − market) the agent will act on. Below this, PASS.5.0%
Max allocation?Kelly-sized cap per decision. Even if Kelly says 10%, the agent won’t exceed this.3.0%
Tail-loss limit?If simulated loss probability exceeds this, the agent PASSes regardless of edge.75.0%
Monte Carlo paths?Seeded simulation count. Higher = tighter intervals, slower dry-run.10,000
No execution · no wallet

Run a dry-run to see the verdict.

Mandate ControlAgent offline
replay clock
VPS telemetry · replay-only

Autonomous Historical Lab

Persistent agent replays evidence in time order; proof revealed after settlement.

Autonomous decision ledger

Every recommendation has a receipt.

Each receipt binds evidence, policy, simulation, and a recomputable hash — a refusal is as auditable as a trade.

Observe

Polymarket + Kalshi

Decide

policy gates + seeded Monte Carlo

Prove

hash-bound receipt, replayable

Loading decision receipts…

Enable Autopilot execution →

Visit Labs for Autopilot & Builder