Rasa

Rasa is an enterprise-grade conversational AI platform for building, deploying, and managing AI agents across text and voice channels. Originally established as an open-source NLU framework, Rasa has evolved into a full platform that combines the flexibility of large language models with deterministic business logic — a hybrid architecture the company calls CALM (Conversational AI with Language Models).

At its core, Rasa lets engineering teams define structured conversation flows, embed business rules, and wire up backend systems, while still leveraging LLMs for natural language understanding and generative responses. This makes it well-suited for regulated industries where conversation behavior must be auditable and predictable — finance, healthcare, insurance, and government are all explicitly targeted verticals.

The platform runs on your own infrastructure, which distinguishes it sharply from cloud-native SaaS alternatives like Dialogflow (Google) or Amazon Lex. Teams that cannot route customer data through third-party cloud services — due to compliance requirements, data residency rules, or internal security policy — tend to find this a decisive advantage. Rasa also supports any LLM backend, so organizations are not locked into a single model provider.

The architecture is modular. NLU handles intent classification and entity extraction for teams that want to start with structured rules. Enterprise RAG layers in real-time retrieval so agents can answer questions from internal knowledge bases without hallucinating. The voice module adds real-time speech infrastructure with low-latency characteristics suited to call center environments. An orchestration layer coordinates multiple agents and tools across a single conversation.

Rasa also supports the Model Context Protocol (MCP), giving agents a standardized interface for connecting to external APIs and tools — aligning with the broader ecosystem direction around interoperable agentic systems.

Compared to Botpress or Microsoft Bot Framework, Rasa offers more fine-grained control over conversation logic and a stronger story around on-premise deployment. Compared to low-code platforms like Voiceflow, Rasa requires more engineering investment but provides correspondingly more flexibility. It sits closer to the developer end of the build-vs-configure spectrum.

Rasa offers a Developer Edition at no cost, making it accessible for prototyping and evaluation before committing to an enterprise contract. The open-source roots mean a substantial community and ecosystem of tutorials, integrations, and third-party resources exist alongside the commercial platform.

Enterprises like Autodesk, BNP Paribas, and Swisscom are listed as customers, suggesting the platform handles production workloads at meaningful scale. Professional services are available for teams that need deployment support or accelerated implementation.