Rasa

Rasa is an open-source-rooted platform for building enterprise conversational AI agents across chat, voice, and messaging channels. Its CALM framework combines LLM flexibility with deterministic logic and built-in recovery patterns, aiming for reliability that a purely LLM-driven bot often lacks. Who it's for: engineering teams building production customer support, internal helpdesk, or sales conversation automation who need predictable behavior, not just plausible-sounding responses, and who want full access to prompts and policies rather than a closed black-box platform. Strengths: on-premises and private-cloud deployment options give real control over sensitive conversation data, and the open framework means no vendor lock-in on the underlying logic. Real-time observability lets teams monitor and improve agent behavior after launch, not just at build time. Weaknesses and limitations: the deterministic-logic-plus-LLM approach has a steeper setup curve than a simple prompt-and-go chatbot builder — it's built for reliability at the cost of some simplicity. Exact paid pricing isn't published; a demo/pricing-page contact is required. Real-world scenario: a support team needs a chatbot that reliably follows a specific refund or account-verification process without going off-script, something a pure LLM chatbot struggles to guarantee.