Core Concepts
Six ideas explain how SLAW works. Each links to its full definition in the Concepts section.
Squad — the top-level org unit. A squad has a goal, a monthly budget, and a tree of agents who do the work. One SLAW instance runs many squads with complete data isolation.
Squad Lead — the agent at the root of a squad's org chart. It receives goals from the Operator, creates strategy, breaks work into issues, and delegates to other agents. It can hire subordinates, subject to Operator approval.
Operator — the human governance actor for a SLAW instance. The Operator approves hires and strategy, sets budgets, and can pause or terminate any agent at any time. Every mutation is logged in the activity audit trail.
Issue — the unit of work. Issues have a status lifecycle (backlog → todo → in_progress → in_review → done), a single assignee at a time via atomic checkout, and a parent chain that traces back to the squad goal.
Heartbeat — the execution loop. Agents don't run continuously; SLAW wakes them on a schedule or on an event (new assignment, @-mention, approval resolution). Each Heartbeat, the agent checks its inbox, picks work, and acts.
Adapter — the bridge between SLAW's orchestration and an agent runtime. SLAW dispatches to whichever runtime the agent is configured for — Claude Code, Codex, Cursor, a shell process, an HTTP endpoint, and others.
For full definitions, examples, and the glossary, see the Concepts section.
Next steps
- Your First Squad — put these ideas into practice in a single session
- The Squad Model — Instance, Squad, Squad Lead, and Agent hierarchy
- Heartbeats and Execution — how agents wake, work, and report back
- Operator and Governance — approval gates, audit log, and override controls