Firewall — a policy gate for agent value flow
Guard answers “is this target safe?”. The Firewall answers “is this agent allowed to do this, right now, within budget, given everything it has already done?” Spend caps, allow/deny, drain protection, anomaly detection and a kill switch — every decision audited.
What it's for
Think corporate-card controls, but for AI agents that hold a wallet. You give an agent spend limits, approval flows and an expense log — and Warden enforces them on every payment and transaction, in real time.
- · An agent is tricked into an unlimited approval → wallet drained
- · A buggy agent spends $10,000/hour instead of $100
- · A regulator asks for proof — you have no record
- · Spend caps stop the runaway agent at $100
- · Unlimited approvals & sanctioned addresses are blocked automatically
- · Every decision is logged — compliance you can prove
- · A kill switch pauses a misbehaving agent instantly
For teams running a fleet of agents — trading bots, agent platforms, DeFi automation, funds. Losing this means a drained wallet, so it's where the real value (and pricing) lives.
Live playground
demo sandboxA shared, throwaway demo agent — not a real dashboard. Flip the policy, fire an action, watch allow · hold · deny and the budget move in real time. Real customers get their own isolated agent, key, budget and audit.
For agents — API
curl -X POST https://warden402.xyz/api/firewall \
-H "content-type: application/json" \
-d '{"action":{"kind":"tx","to":"0x..","from":"0x..","calldata":"0x095ea7b3.."}}'
# → { result: { decision: "deny", reasons: ["UNLIMITED_APPROVAL"], ... } }