DocsAugust
2026
Build with Lloyal
Lloyal moves application control from the edges of a model call into the inference trajectory itself. These guides progress from a working application, through the execution model beneath it, to adaptive policy and direct inference patterns.
One curriculum, four depths
Build your first harness →
Use harness.dev to create an application, begin a Session, inspect its procedure, change its topology, and continue from accepted state.
Thinking in Lloyal →
Learn how owned lifetimes, live inference state, Agents, Tools, orchestration, finality, and continuity fit together.
How Agents Know When to Stop →
Learn how policy, context pressure, retries, recovery, wind-down, and completion govern a running AgentPool.
Advanced patterns →
Program the inference trajectory directly with verifier-before-commit generation, diverse tree search, expert-state synthesis, verified continuation, and replay.
Recommended learning path
One harness, different placements
Where a harness runs
Understand how one harness contract spans application surfaces, local and shared-residency execution, process boundaries, transports, hardware, and operators—using the familiar Rails architecture as the bridge.