- What is builtbyberry/laravel-swarm used for?
- It is a Laravel package for building multi-agent AI workflows. You define agents, tools, and handoffs as PHP classes and coordinate them through Laravel-native configuration, events, and console tooling. It is built on Laravel AI and is intended for applications that can own an AI orchestration layer.
- How do I install laravel-swarm in a Laravel app?
- Install it with Composer as builtbyberry/laravel-swarm and review the package README for any required service provider or config publishing steps. Because it is pre-1.0, pin the version you validate and check composer.json for exact Laravel and PHP constraints. Plan a controlled migration rather than treating upgrades as routine patches.
- Which Laravel and PHP versions does laravel-swarm support?
- The provided evaluation does not list exact Laravel or PHP version constraints, so check the package composer.json and README before adopting it. The package is Laravel-native and uses config, service container, events, and console commands. Treat version support as something to verify and pin in your own project.
- Is Laravel Swarm production ready?
- It is best treated as pilot-grade for v0.28.1, not a routine production patch. The package is still maturing, and v0.28.1 includes behavior changes rather than only additive fixes. Adopt it where your team can own AI orchestration semantics, instrument memory and tool behavior, and absorb pre-1.0 API churn.
- What are agents, tools, and handoffs in Laravel Swarm?
- Agents are the AI workers you define as PHP classes. Tools are capabilities the agents can invoke, and handoffs let one agent pass work to another. Laravel Swarm coordinates these through route-plan schemas, hierarchical planning, persistent swarm memory, and replay boundaries.
- How does swarm memory work in laravel-swarm?
- Laravel Swarm provides persistent swarm memory, including Remember writes, MemoryCapturePolicy behavior, replay modes, and checkpoints. In v0.28.1, Remember writes that store nothing or are skipped by capture policy now report as failed, and a skipped write at save time fails the step. Review any dashboards, alerts, or logic that interprets memory-write success.
- What is agent-scope memory and why is it a multi-tenancy concern?
- Agent-scope memory is an opt-in feature enabled through the swarm.memory.tools.agent_scope config key plus swarmMemoryTools(agentScope: true) or a custom forAgent() binding. When enabled, it is keyed by agent class and shared across runs and tenants of that class. That makes tenant isolation, retention, and cross-run sharing explicit review items before enabling it.
- What behavior changes should I check before upgrading to v0.28.1?
- Review Remember outcomes that move from succeeded to failed, Skip writes that now fail the step, and frozen_view retry memory writes that persist under the default replay mode with no switch. Also check RoutePlanSchema::worker() consumers because it now requires with_outputs and declares next as required nullable. Snapshot tests and any nativeResult->tools or SwarmToolResult consumers should be revalidated.
- How should I test laravel-swarm before using it in production?
- Use a live-model regression suite because v0.28.1 is a post-ship fix release against live-model behavior. Pin the package and gate upgrades behind those tests. Also review tests, dashboards, alerts, and logic that interpret SwarmToolResult events, nativeResult->tools, or memory-write success.
- How do I create memory tools with make:memory-tool?
- Use the make:memory-tool command, including the --scope=agent option for agent-scope memory tools. Existing agent() overrides keep working but are not governed by the config key, and any null-returning placeholder must be removed before binding with forAgent(). Enable agent scope with the swarm.memory.tools.agent_scope config key and explicit swarmMemoryTools(agentScope: true).