Guide
Multi-agent workflows in Ruby
Not every problem is one mega-prompt. Split research, decision, and action agents when responsibilities differ.
Updated 2026-07-28 · Rails Agent · Tiny Bubble Company
Direct answer
Multi-agent workflows in Rails compose typed agents—knowledge, workflow, operations, monitoring—with explicit handoffs. Rails Agent’s taxonomy and directory model keep each agent’s tools and prompts scoped.
When to split agents
- Different tool permissions
- Different models or cost profiles
- Clear human approval boundaries
- Independent eval suites
Anti-pattern
Spawning agents without ownership of state creates race conditions. Pass structured handoff payloads and keep a single writer for critical records.
Frequently asked questions
How many agents should a workflow use?
As few as possible. Split when permissions, evals, or models diverge—not for novelty.
How do taxonomy types help?
Knowledge, Workflow, Operations, and Monitoring map to different base classes and expectations; see /docs/agent-types.
Ship production Rails AI agents
Install rails-agent-stack, open /agents, and go from scaffold to deploy with Playbooks, memory, guardrails, and monitoring included.