Agentic Trading
Also: AI trading agents · autonomous trading
Agentic trading is the use of LLM-driven agents to generate, evolve, and operate trading strategies — not just to produce one signal among many, but to run the loop end to end under human-set policy.
Most “AI trading” bolts a model onto a rules engine: the model emits a score, hardcoded logic decides the trade. Agentic trading hands the whole loop to an agent — it forms a hypothesis, sizes a position, places the order, watches the fill, and revises. The model isn’t one input; it’s the operator.
That only works if autonomy is bounded. An agent with unlimited authority and your private key is a liability, not an edge. The practical design separates what the agent may decide from what it may execute without you.
On Lore this is the Plan / Auto / Bypass spectrum. In Plan, the agent proposes and you approve every action. In Auto, it executes freely inside a policy envelope — position caps, allowed markets, max leverage — and anything outside the envelope bounces back to you. In Bypass, you’ve granted scoped full trust. The agent is a co-signer in a non-custodial MPC wallet, so it can act but never owns the keys, and you can revoke authority at any time.
- agents
- automation
- policy
Research source: rSwarm research library →