A Rails engine for orchestrating AI agents — with ActiveRecord persistence, background-job execution, generators, a web dashboard, built on rcrewai 0.10.
A batteries-included engine for AI agent orchestration in Rails.
Persist crews, agents, tasks, and executions directly in your database.
Runs through ActiveJob — works with Sidekiq, Resque, Delayed Job, or any adapter.
Scaffold new crews and agents from the command line in seconds.
Monitor executions, view logs, and manage crews through a built-in UI.
OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, Ollama, AWS Bedrock, Snowflake Cortex, and any OpenAI-compatible endpoint.
Pre-built tools for ActiveRecord, ActionMailer, the Rails cache, and more.
rcrewai features through 0.9, surfaced through ActiveRecord, on rcrewai 0.10.
Reasoning passes, per-agent LLM overrides, rate limiting, context-window trimming, and cognitive memory.
Structured output schemas, guardrails, file output, and multimodal image attachments.
before/after kickoff hooks, planning, the consensual process, and batch execution.
String, file, PDF, CSV, and URL sources attached to agents or crews.
Persist and resume rcrewai Flow state through the database; track each run.
Semantic recall with a pluggable embedder and SQLite-backed store.
Span-level traces of every run, with timings, tokens, cost, and a live view.
Resume an interrupted run without re-running the tasks it already finished.
Hooks around every LLM request, plus Bedrock, Snowflake, and OpenAI-compatible endpoints.
A turn's tool calls run in parallel, with a per-agent opt-out.