The most advanced agentic platform for Ruby on Rails.
Build Rails AI agents in your repo. Test, deploy, and monitor at /agents.

AGENTS.md · Getting started · Coding agents · Cloud

Rails Agent is the fullstack Rails agent framework for production Ruby on Rails agents. Define instructions, tools, skills, packages, knowledge, channels, memory, and evals as files in your app — then use the mounted /agents dashboard to connect BYOK credentials, attach guardrails, ship channels, and observe cost.

If you are searching for a Rails AI agents stack, an ActiveAgent alternative, or a RubyLLM alternative, Rails Agent is built for the full lifecycle: author in Git, run in the cloud, integrate responses into your product.

Website: rails-agent.com

Rails Agent is a better alternative to library-only LLM wrappers when you need agents that ship with your Rails app — not just chat completions.

Positioning in one line: most advanced agentic platform for Ruby on Rails — git-native agents with cloud runtime, integrations, and production controls.

• Playbooks — proven scaffolds (instructions + tools + skills) you customize locally
• Library — share tools, skills, packages, knowledge, and connectors across agents (app/agents_library/ + imports.yml)
• Budget — per-agent / policy spend limits alongside usage monitoring
• Monitor — runs, traces, evals, and cost in /agents
• BYOK — attach OpenAI, Anthropic, and other provider keys in cloud; reference them in Ruby with no secrets in Git
• Open-Wire — Slack and Teams channel transport over open-wire/1
• Composio email — email connectors for inbox workflows
• Evals — golden cases under evals/ before deploy
• Deploy — rails-agents sync → deploy → production traffic
• Coding-agent docs — Cursor, Codex, and Claude Code follow AGENTS.md and Coding agents

# Gemfile
gem "rails-agent-stack", "~> 0.1"

bundle install
bin/rails generate rails_agents:install
bin/dev   # → http://localhost:3000/agents

Connect your company workspace at /agents, then scaffold a database Knowledge agent:

bin/rails generate rails_agents:agent store_assistant --type knowledge --database

Zeitwerk: fresh installs add config/initializers/rails_agents_autoload.rb so app/agents/ and app/agents_library/ are agent assets, not application namespaces.

External coding agents (Cursor, Claude Code, Codex): start with docs/AGENTS.md.

bundle exec rails-agents run store_assistant "Where is order ORD-DEMO-1001?"
bundle exec rails-agents sync store_assistant
bundle exec rails-agents deploy store_assistant
bundle exec rails-agents logs store_assistant

class StoreAssistant < RailsAgents::KnowledgeAgent
  model :gpt_5_mini, provider: :openai, credential: :company_openai
  memory :conversation, provider: :mem0, recall: 5
  knowledge_from "knowledge/**/*"

  tool :lookup_order do |order_number:|
    Order.find_by(number: order_number)&.as_json(only: %i[number status total_cents])
  end
end

:company_openai is a cloud credential reference — not a local env var. Enable managed memory from Build → Memory. Pass a stable session_id to .run so recall stays scoped per product user.

app/agents/store_assistant/
├── agent.rb            # RailsAgents::KnowledgeAgent subclass
├── prompt.md           # System prompt
├── imports.yml         # Optional workspace Library attachments
├── tools/              # Optional extracted tools
├── skills/             # Composable behaviors
├── packages/           # Registry-installed capabilities
├── memory.rb           # Memory config
├── knowledge/          # RAG files
├── connectors/         # External connection manifests
├── channels/           # Slack, Teams, web, API, …
└── evals/              # Smoke / regression cases

app/agents_library/
├── tools/
├── skills/
├── packages/
├── knowledge/
└── connectors/

Keep one-off work inside the agent. Move stable capabilities to app/agents_library/ when another agent should reuse them, then reference via imports.yml. rails-agents sync includes imported Library files in the runtime bundle without duplicating shared source in Git.

Generator: --type knowledge|workflow|operations|monitoring and --database for DB-connected Knowledge scaffolds.

Integrate responses: every run returns output_text (Markdown), optional output_data, and typed items — same shape for every LLM. Guide: rails-agent.com/docs/run-response.

# config/initializers/rails_agents.rb
RailsAgents.configure do |config|
  config.api_key = ENV["RAILS_AGENTS_API_KEY"]       # platform key
  config.project_id = ENV["RAILS_AGENTS_PROJECT_ID"]
end

Provider keys live in the cloud dashboard → Model credentials (BYOK). Never commit them.

Details: rails-agent.com/pricing

bundle install
bundle exec rspec

Local path testing: sibling app rails-agents-boilerplate.

MIT — MIT-LICENSE

rails-agent.com · AGENTS.md · Getting started