SLAW vs Alternatives
SLAW sits in a specific position: it's not a task manager, not a raw agent terminal, and not an agent framework. Understanding where it differs helps you pick the right tool.
SLAW vs task managers (Linear, Jira, GitHub Issues)
Task managers track work done by humans. They have no concept of running an agent, issuing a heartbeat, enforcing a budget cap, or approving a hire. You can use a task manager alongside SLAW (SLAW ships a Jira sync plugin), but you can't use one instead.
What SLAW adds on top of a task board: agent assignment and lifecycle, heartbeat scheduling, atomic checkout (one agent owns a task at a time), budget enforcement, delegation hierarchy, approval gates, and an audit log of every agent action.
SLAW vs running agents directly (terminals, scripts)
Running Claude Code or Codex directly in a terminal is the right tool for a one-off task with one agent. It breaks down when you want:
- Multiple agents coordinating on shared work
- Tasks that survive a terminal session
- Budget caps that actually stop an agent
- An audit trail that persists
- Human approval gates before sensitive actions
- A way to monitor what's running from a phone or a dashboard
SLAW wraps the agents you're already using — it calls Claude Code, Codex, Cursor, or any HTTP endpoint — but adds the coordination, governance, and persistence layer on top.
SLAW vs agent frameworks (LangGraph, CrewAI, AutoGen)
Agent frameworks are code-first. They define agent behavior in Python or TypeScript, orchestrate multi-agent graphs in-process, and typically run as a single program. They're the right choice if you're building an AI product and you own the code that defines agent behavior.
SLAW is operations-first. It orchestrates agents that already exist — Claude Code, Codex, Cursor — as external processes. The "code" that defines agent behavior is the adapter config and the agent's capability description, not a Python class. SLAW is what you reach for when you want to run a squad, not build an agent.
The distinction matters for data model, too. Frameworks work in memory for the duration of a run; SLAW is persistent-first — every issue, comment, and cost event is stored in Postgres and survives restarts.
Summary
| SLAW | Task manager | Terminal / script | Agent framework | |
|---|---|---|---|---|
| Runs agents | ✓ | ✗ | ✓ (one) | ✓ |
| Multi-agent coordination | ✓ | ✗ | Manual | ✓ |
| Budget enforcement | ✓ | ✗ | ✗ | ✗ |
| Persistent task state | ✓ | ✓ | ✗ | ✗ |
| Operator approval gates | ✓ | ✗ | ✗ | ✗ |
| Local data sovereignty | ✓ | ✗ (cloud) | ✓ | Depends |
| Bring your own agent runtime | ✓ | N/A | ✓ | ✗ |
| No code required | ✓ | ✓ | ✗ | ✗ |
Next steps
- What is SLAW? — The full picture
- Quickstart — Try it in minutes
- How agents work — The adapter model in detail