Active Agent

Active Agent is a framework designed for integrating AI agent functionality into Ruby on Rails applications. It enables developers to build agents that function as controllers, using familiar Rails conventions such as actions, callbacks, and views. The platform is aimed at Rails developers seeking to incorporate AI-driven features into their products while maintaining a native development experience.

The framework supports composing prompts for AI models in a manner similar to how Action Mailer works for email, allowing developers to define prompts and actions within agent classes. Active Agent offers compatibility with multiple large language model providers, including OpenAI, Anthropic, and Ollama, and switching between these providers can be accomplished with minimal code changes. It includes features for resilient agent operation, such as automatic retries, exponential backoff, and graceful failure handling. Real-time streaming of responses is supported, enabling agents to provide incremental outputs as they process data.

For structured data needs, Active Agent allows responses to be formatted according to JSON schemas, ensuring typed and instruction-following outputs. Tool calling is another capability, where AI agents can invoke Ruby methods to fetch data, perform actions, or make decisions based on AI-generated instructions. The framework also supports the Model Context Protocol (MCP), which allows agents to connect to various MCP servers, including integrations with file systems and GitHub.

Active Agent integrates with ActiveJob for background job processing, making it possible to run agent tasks asynchronously and at scale. Observability features are built in, providing tracing, measurement, and evaluation of agent interactions in production environments. Developers can trace prompts, reasoning, model calls, tool use, and token costs, and evaluate outputs using LLM-based or rule-based checks.

The framework is delivered as a set of Ruby gems for use within Rails projects.