Where your agents actually run — under the controls you set.
SLAW is the self-hosted runtime each team runs under the control plane. It executes AI agents inside an org chart, central budgets, and approval gates — bringing any agent with an adapter (Claude, Codex, Cursor) under one set of controls, with every action logged and config and secrets never leaving the machine.
Self-hosted only · no hosted version · governed by the control plane
How it works
A control plane that orchestrates — it doesn't run your agents.
On each heartbeat, SLAW looks up an agent's adapter and config, hands it the execution context, and the agent runs wherever it runs and returns to the local server. SLAW captures the result, parses usage and cost, and moves the work forward — all inside your environment.
- The heartbeat is the loop. Agents wake on a schedule, check work, and act — every cycle metered, gated, and logged.
- Atomic checkout. Only one agent owns a task at a time; a simultaneous claim returns 409.
- Embedded by default. Bundled PostgreSQL, local data dir, dashboard at
localhost:3100. No external services to wire up.
The squad model
A real org chart for your agents.
The hierarchy is Instance → Operator → Squad → Squad Lead → Agents. One instance runs many squads with complete data isolation — one control plane for a whole portfolio.
SLAW is the orchestration half of the Agentic Control Plane in the Enterprise Agentic Reference Architecture. Botfather is the governance half.
Operator
The human governance actor for the instance. Approves hires, sets budgets, overrides strategy, and can pause or terminate any agent. Owns the audit log.
Squad & Squad Lead
A squad is the top-level org unit with a goal, a budget, and a task hierarchy. The Squad Lead drafts strategy, breaks goals into tasks, and is the primary delegator.
Agents
Every employee is an AI agent in a strict tree — one manager each. Each carries an adapter, a role, capabilities, a monthly budget, and a live status.
Bring your own agent
Adapters bridge SLAW to any runtime.
SLAW orchestrates; the adapter executes. Standardise the agents your teams run behind one set of controls — without locking into a single provider.
When you need it
Reach for SLAW when agents stop being an experiment.
- Teams are already running agents on their own, off the books and off the budget.
- You need agents from different providers behind one consistent set of controls.
- Agents act on production code, data, and tickets — and that has to be auditable.
- Spend has to be capped and enforced, not discovered at month end.
- Code, data, and secrets must stay inside your environment — self-hosted, no exceptions.
Run agents you can actually control.
Deploy SLAW on your own infrastructure, then bring it under the control plane to govern your enrolled teams from one place.