Rasa

Rasa provides a developer platform for building, deploying and managing enterprise AI agents for chat and voice. Its stated approach is to extend large language models with an organization's business logic so that agents remain controllable and predictable across large conversation volumes, using LLMs to interpret user language rather than to infer business rules or drive the flow of a conversation. The company describes this method as CALM (Conversational AI with Language Models), positioned as a combination of LLM flexibility with the determinism of traditional NLU-based chatbots.

The platform is delivered through several components. Rasa Pro is the commercial, pro-code offering built around security, observability and scale on top of the open source framework; Rasa Studio is a lower-code or no-code interface for designing and refining agents; and Rasa X/Enterprise historically served as a low-code review interface used alongside Rasa Pro. Product documentation also covers a Rasa Copilot for accelerated agent building, Rasa-specific MCP tools for IDEs such as Cursor, Claude Code, VS Code and JetBrains, a browser playground, and a Rasa University certification course.

Advertised platform capabilities include a patented dialogue management system that orchestrates autonomous reasoning, guided workflows and shared conversational memory; multi-agent orchestration with shared state and task handoff between domain-specific skills; on-premises, private cloud and air-gapped deployment options; support for customer-chosen LLMs and infrastructure with access to prompts, policies and underlying code; real-time voice and IVR features such as streaming speech recognition, barge-in, turn-taking and noise handling; and observability tooling for conversation tracing, evaluation and production monitoring.

Business model

Open core: an open source conversational AI framework underpins commercial licensed products, notably Rasa Pro, which requires a paid license, alongside Rasa Studio and the broader Rasa Platform sold to enterprises. Deployment can occur on customer infrastructure, private cloud or offline environments.

Software licensing and subscription revenue from enterprise products layered on the open source framework; Rasa Pro capabilities are gated behind a commercial license.

Traction

Rasa says its open source framework has been downloaded more than 50 million times as of early 2024, with legacy documentation citing over 25 million downloads for Rasa Open Source. The company reported roughly doubling annual recurring revenue in 2023 versus 2022. Disclosed customers include two of the world's three largest banks, two of the largest U.S. banks, American Express and Deutsche Telekom.

Latest developments

Recent materials describe the Rasa Platform with multi-agent orchestration, sub-agents connected to MCP servers, voice and IVR support, and Rasa Copilot and Rasa Tools (MCP) for building agents in IDEs from version 3.16 onward. The site also cites the company's Strong Performer placement in the Q2 2026 Forrester Wave for conversational AI platforms for customer service.