Concepts

A factory bottleneck is the stage or resource that caps overall you-agent-factory throughput even when other stages have spare capacity.

A factory planner checklist is a live outcomes and workstream board that shows done, active, ready-next, and intentionally held work.

Compaction shrinks retained context—history, transcripts, or stream windows—so long-running agent or factory work can continue within limits.

A harness is the agent runtime the factory drives, such as a coding agent under a shared factory process.

A factory loop keeps iterating under you-agent-factory until the goal is done.

Model Context Protocol: the host↔server protocol that exposes named tools, including Factory Session tools via you mcp serve.

Reusable instruction packages a harness or coding agent can load to specialize behavior for a task class.

Statistical process control graphs, also called control charts, plot a factory process metric over time with a center line and control limits so readers can tell routine variation from unusual signals.

The you-agent-factory task queue is the backlog of submitted work waiting in work-type states until workstations consume it.

Thinking is the deliberative reasoning a harness or model may produce before or alongside tool use and final answers, including thinking controls that trade quality against latency and cost.

Model-inference tokens: units of model input and output used for context limits and cost accounting when running harnesses, workers, or inference.

A named callable capability an agent or Model Context Protocol (MCP) host can use while doing factory work.

Tool calling is the model-inference behavior of selecting and invoking named tools during an agent or model turn, gated by worker tool policy such as agentTools.

A worktree is an isolated checkout or workspace the factory uses so agent lanes can run without colliding on the same tree.