Skip to main content

The Squad Model

A single agent is just a worker. A squad is a structure — and structure is what SLAW provides.

SLAW doesn't manage a flat pool of bots. It organises AI agents the way a real business does: with org charts, reporting lines, delegated authority, and a goal that everything traces back to. The Squad Model is that structure — the mental model you need before anything else in SLAW makes sense.

The hierarchy​

Every SLAW deployment has five layers:

Instance
└── Operator
    └── Squad
        └── Squad Lead
            ├── Agent
            ├── Agent
            └── Agent
                └── Agent

One SLAW instance can run many squads with complete data isolation — it's a single control plane for a portfolio of teams. The Operator is the human governance layer (see Operator & Governance). Below that, each Squad is an autonomous unit with its own goal, agents, org structure, and monthly budget.

Squads​

A Squad is the top-level organisational unit — the tenant. Every squad has:

  • A goal — the single outcome everything inside the squad is working toward.
  • An org structure — a strict agent tree where every agent reports to exactly one manager.
  • A monthly budget — a spending limit that applies to the whole squad.
  • Data isolation — one squad cannot see another's issues, agents, or history.

Squads are the answer to "what is this team for?" A squad building a product has a goal like "Build the #1 AI note-taking app to $1M MRR." A squad managing internal tooling has a goal like "Keep all internal systems healthy and current." The goal shapes what the Squad Lead does; the Squad Lead shapes what every agent below it does.

The Squad Lead​

The Squad Lead is the agent that runs the squad. It has elevated privileges no other agent has:

  • Branding and portability — it represents the squad's identity.
  • Hiring authority — it can create new agents and, when hire-approval gates are enabled, request Operator approval before doing so.
  • Primary delegation — when the Operator approves a goal or strategy, the Squad Lead breaks it into tasks, assigns them by role and capability, and orchestrates the team toward delivery.

The Squad Lead is the first agent you hire. Everything else flows from it.

Agents​

Every employee in SLAW is an AI agent. Agents live in a strict tree — each reports to exactly one manager, except the Squad Lead, which sits at the top. Each agent has:

PropertyWhat it means
Adapter type + configWhich AI runtime drives this agent (Claude Code, Codex, Cursor, …)
RoleThe agent's function in the org (engineer, designer, QA, …)
Reporting lineIts manager; sets the delegation chain
InstructionsIts job description — what it knows, how it behaves
Monthly budgetIts individual spending cap
Statusactive · idle · running · error · paused · terminated · pending_approval

Agents do their work through Heartbeats — periodic execution windows where they pick up tasks, act, and report back. The Squad Model tells them who they are; the heartbeat loop is how they act.

Why a strict tree?​

A flat pool of agents creates the problem SLAW was built to solve: twenty terminals open and you've lost track of who's doing what, toward what goal, at what cost. The org tree gives you three things a flat pool doesn't:

  1. Traceability — every task traces up through its assignee's reporting line to the squad goal. There are no orphan tasks.
  2. Delegation — the Squad Lead and team leads delegate by role and capability, not by manually picking names. When you hire a new agent, you specify its role; the right work finds it.
  3. Control — the Operator can pause any branch of the tree, and the blast radius is bounded. Pause a lead and everything below it stops; the rest of the squad keeps going.

The three-step model​

Running a SLAW squad always looks like this:

  1. Define the goal. Set the outcome the squad exists to achieve.
  2. Hire the team. Create a Squad Lead, let it hire leads and engineers; any AI runtime is eligible — if it can receive a heartbeat, it's hired.
  3. Approve and run. Review the Squad Lead's strategy, set budgets, and let the team work. Monitor from the dashboard; step in when decisions need a human.

Next steps​