Skip to content
Lootrunners
Small blast radius

Assume an agent can fail. Limit what failure can cost.

Lootrunners separates agent capability from financial authority. Application policy defines the rich boundary — purpose, vendors, Rules, approval tiers. Independent provider or wallet controls enforce critical ceilings where the payment rail supports it. Sensitive changes stay human-controlled and auditable.

S·01
Scoped access

Connecting a service does not give every agent access. Resources can be granted to one agent or one Run. Credentials stay behind the connector; the model receives only the capability it needs.

S·02
Bounded authority

Every autonomous purchase is evaluated against a Mandate and the shared Rules that apply.

S·03
Independent limits

Critical ceilings can be mirrored into supported provider or wallet controls and verified separately from application policy.

S·04
No self-escalation

Agents cannot widen their own Mandates, approve their own exceptions, grant themselves Resource access, or unfreeze spending.

S·05
Verifiable freeze

One action blocks new autonomous spend and revokes supported external permissions — then shows you exactly which revocations succeeded.

S·06
Explainable decisions

AUTO, ASK, and BLOCK decisions keep the rule, evidence, and actor behind them. Workspaces using the Rules library see the governing Rule and version cited.

The signature moment

"You are giving Research Runner permission to spend up to $50 USDC every 24 hours until August 20. Lootrunners can apply stricter rules, but it cannot use this permission to exceed that limit."

No wallet signing prompt ever appears before this plain-English explanation.

Operational safeguards
Forced row-level tenant isolation and least-privilege database roles
Step-up authentication for high-consequence authority changes
Agent and model environments do not receive treasury signing credentials
Resource credentials stay behind scoped connectors
No raw private keys in application config
Deterministic policy evaluation for financial decisions
Idempotent payment execution — unknown settlement state is never blindly retried
Freeze reports partial revocation honestly
AI explanations cannot change authority

Start risk-free on testnet.

Test agent payments with fake funds before giving them access to real money.

Try the sandbox