Rasa
Open-source conversational AI framework with full data control
Overview
Rasa is an open-source ML framework for on-premises conversational AI, built for regulated industries that can't send chat data to third-party servers. Its Pro tier adds CALM's LLM-native dialogue management, plus the no-code Rasa Studio builder that replaced the retired Rasa X.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- CALM dialogue engine
- 1,000 conversations/mo
- Community support
- Rasa Pro + Studio
- Enterprise search & RAG
- SSO & role-based access
- 24/7 premium support
- Dedicated CSM
- Enhanced security
Use cases
What you can produce with Rasa
- Intent classification and entity recognition that engineers with Python and NLP depth rate highly — serious chatbot infrastructure rather than a toy
- Full data control, with the framework running on your own systems
- CALM: LLM-handled understanding with the business logic still locked in code, Rasa's deliberate answer to letting a model freewheel through a conversation
- A Developer Edition that is free and genuinely usable for smaller bots
- ⚠️ Real ops complexity. This is not drag-and-drop, the action server and the Rasa server run as separate processes that have to be wired together correctly, and debugging gets messy — teams without in-house NLP or Python depth tend to bounce toward Voiceflow or Dialogflow
- ⚠️ A sales call past 1,000 conversations a month: there is no self-serve paid tier and no public number to budget against, and Rasa X's conversation-review functionality now sits inside Rasa Studio, bundled into the paid Business and Enterprise tiers
Rasa is the established open-source framework for building custom conversational assistants, giving developers full control over dialogue logic, NLU, and deployment. Its CALM approach blends LLMs for more natural flows while keeping business logic deterministic in code — a meaningful edge for regulated industries needing on-prem, data-sovereign assistants.
It's developer-heavy: building and maintaining a Rasa bot takes real engineering investment versus drag-and-drop builders, and the Business/Enterprise tiers are sales-priced with no public rate card. Rasa X was retired in 2025, folded into Rasa Studio's no-code builder. For ASEAN teams needing local data control and multilingual custom assistants, it's attractive — provided they have engineering capacity to match.
What people say
4.7 out of 5 on G2, but that's from just 37 reviews — a small sample for a framework this widely deployed. On TrustRadius and Capterra the pattern repeats: engineers who already know Python and NLP concepts rate Rasa highly for intent classification and entity recognition, calling it one of the better frameworks for building serious chatbot infrastructure rather than a toy.
The complaints cluster around ops complexity. Reviewers flag a steep learning curve — Rasa isn't drag-and-drop, and debugging gets messy because the action server and the Rasa server run as separate processes that have to be wired together correctly. Teams without in-house NLP or Python depth tend to bounce off it toward hosted alternatives like Voiceflow or Dialogflow.
The bigger 2026 story is CALM, Rasa's pivot toward LLM-handled understanding with business logic still locked in code — a deliberate answer to the reliability worries that come with letting an LLM freewheel through a conversation. Rasa X, the old conversation-review tool, was sunset in 2025; that functionality now lives inside Rasa Studio, bundled into the paid Business and Enterprise tiers.
Pricing is the other sticking point: the Developer Edition is free and genuinely usable for smaller bots, but anything past 1,000 conversations a month forces a call with sales — there's no self-serve paid tier and no public number to budget against.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- Rasa is written in Python and has been downloaded over 25 million times, making it one of the most widely used open-source NLP frameworks in the world.
- The dialogue management system uses machine learning trained on annotated conversations, meaning the bot can handle unexpected user journeys it was not explicitly programmed for.
- Deutsche Telekom deployed Rasa to handle over 100 million customer conversations per year in 24 languages — one of the largest on-premises conversational AI deployments.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Rasa's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Rasa unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Rasa directly →
Spotted something out of date? Suggest an update →
More in Agents & Automation
