# Morgana

A conversational AI framework for .NET that orchestrates specialized agents through an actor-model architecture to classify and resolve customer inquiries with semantic routing and policy guardrails.

Morgana is built on .NET 10 and uses Akka.NET's actor model to create a fault-tolerant, scalable orchestration layer. Each conversation runs through a hierarchy of actors: a conversation manager owning a session's lifecycle, a supervisor coordinating the flow, a guard validating interactions against business policies, a classifier analyzing intent with an LLM, and a router dispatching to the right agent.

Agents are domain specialists that self-register through declarative attributes rather than manual wiring, each inheriting from a base agent class. At startup, Morgana automatically discovers agents across configured assemblies and validates that declared intents and classifier configuration are consistent, failing fast before reaching production. Agents can also expose native or Model Context Protocol and Agent-to-Agent tools.

It is powered by Microsoft's agent framework for LLM integration, with governance enforced through guard rails and policy checks, and supports multi-channel conversations.

Website: https://mdesalvo.github.io/Morgana