Owner vs Agent Authority: Managing Trading Automation Controls
Explore the distinction between owner and agent authority in trading automation, the controls each provides, and best practices for safe delegation.
Produced with automation, then checked by deterministic quality rules and an independent source-grounded review before publication.
- 01Owner authority can sign any transaction, set policy limits, and withdraw funds.
- 02Agent authority is restricted to actions explicitly permitted by the owner.
- 03Trade‑scoped agent keys cannot initiate withdrawals; a separate owner intent is required.
- 04Owner‑defined limits include order size, daily notional, daily loss, and expiry but do not guarantee loss avoidance.
- 05Emergency stops revoke the calling key but do not automatically close positions or cancel allowances.
Owner authority gives the account holder the ability to sign any transaction, define policy limits, and withdraw funds. Agent authority operates under scoped keys that restrict actions to those the owner explicitly permits. Together they form a layered control model for automated trading.
What does owner authority enable?
Owner authority is the ultimate source of power over an account. It can sign all transaction types, including withdrawals, and it defines policy fields such as order size caps, daily notional exposure, daily loss limits, and expiry dates. The owner must monitor these settings to prevent misuse or unexpected market events. Because the owner key controls the withdrawal path, any movement of capital out of the account requires a distinct signed intent from the owner. This separation ensures that even if an agent key is compromised, the attacker cannot directly extract funds.
How is agent authority constrained?
Agent authority is granted through scoped keys that limit the agent to specific trade‑related actions. These keys can enforce order‑size and notional caps, time‑bound validity, and policy‑linked loss thresholds. They cannot initiate withdrawals, and any attempt to exceed the defined scope is rejected by the system. The scope is defined at key creation and can be adjusted only by the owner, preserving a clear boundary between execution and fund movement.
Typical scopes for an agent key
- Trade‑only scope: place, modify, or cancel orders within set size and notional limits.
- Time‑bound scope: expires after a defined period and requires renewal.
- Policy‑linked scope: inherits owner‑defined daily loss or notional caps.
Why can’t an agent key be used for withdrawals?
Withdrawal actions require a separate owner‑signed intent because they move funds out of the controlled account. Even a trade‑scoped agent key with high order limits lacks the authority to initiate a withdrawal, preventing a compromised agent from siphoning assets. The system enforces this rule by checking the signature type at execution time; only a signature derived from the owner key satisfies the withdrawal condition.
What happens during an emergency stop?
An emergency stop revokes the calling key’s authority, halting further actions from that agent. It does not automatically close open positions or cancel existing token allowances; those require explicit owner review and separate actions. The stop provides a rapid containment mechanism while leaving residual exposure for the owner to manage. After a stop, the owner should reconcile the current state, verify open orders, and decide whether to close positions manually or adjust limits.
"Controls are only as effective as the monitoring and response processes that accompany them."
How can owners monitor and adjust controls?
Owners should regularly review runtime status, account state, and market data quality to ensure agents operate within intended limits. Adjusting policy fields such as daily notional caps or loss thresholds can tighten or relax constraints as market conditions evolve. Continuous monitoring is essential because a timeout does not prove an order failed; reconciliation and durable mutation identity confirm outcomes. Owners can also use audit logs to trace each decision made by an agent, providing transparency for post‑trade analysis.
For deeper context see Understanding Owner and Agent Authority for Trading Agents, Owner Authority vs Agent Authority: Understanding the Core Differences, and How Owner and Agent Permissions Differ in Trading Automation.
Frequently asked questions
Owner authority can sign any transaction and set policy limits, while agent authority is limited to actions explicitly permitted by the owner and cannot withdraw funds.
No. Agent keys are created with scoped permissions and cannot be transformed into owner keys; a new owner key must be used for full control.
No. Limits constrain order size and exposure, but market moves can still cause losses within the allowed range, and execution uncertainty remains.
Review open positions, verify token allowances, and decide whether to close positions or adjust limits. The stop only revokes the key, not existing exposures.
Regular review is recommended, especially after significant market events or strategy changes, to ensure limits remain appropriate.
Sources and verification
Product claims in this article were checked against these first-party references. Runtime status remains authoritative for current availability.
- Felix documentationfirst party
- Felix machine referencefirst party
Build with Felix now.
Felix infrastructure is live through MCP and the API. The Felix V1 retail quant-desk private beta is planned for September 22.
AI trading bots continuously process market data, use trained models to create trade signals, and convert those signals into signed orders while enforcing strict risk controls and safety checks.
Choosing the right API is essential for reliable algorithmic trading. This guide details data quality, permission models, latency, error reporting and risk controls to help you evaluate options.