Hello Rasa FAQ

"Hello Rasa" is an interactive playground for developers to prototype AI agents with Rasa. Select a template (banking, telecom, or customer support), get a working agent up and running in minutes, and customize it by prompting or editing code directly. No setup required, just open it and start building.

Open the playground, select one of the three available templates (banking, telecom, or customer support), and start building your agent in the browser. No account creation, no installation, no configuration files to set up. You're building a prototype agent in under five minutes.

No. The playground gets you building immediately with working templates. Learn by doing, ask the copilot questions, and watch the inspector panel to see how things work. When you want to build something new, ask the copilot on the left to generate some improvements or a new scope. You'll understand Rasa's core concepts by getting hands-on and actually building with them.

Three pre-built templates (for now!) to get started: banking (account management, transactions, card services), telecom (plan upgrades, billing, technical support), and general customer support. Each template has only a few flows and, therefore, a limited number of questions it can answer out of the box. Prompt the copilot to generate a few new flows so you can extend its scope.

Reliable AI agents for customer support, banking operations, telecom services, or any conversational workflow where you need natural language understanding combined with predictable business logic. The playground focuses on high-trust use cases where consistency matters.

Absolutely, that is exactly what we'd love for you to do. Prompt the copilot to add features, change behavior, or fix issues. It generates code you paste in. Alternatively, you can edit the code in the files yourself. You have full access to everything. The templates are starting points, not constraints.

A built-in agent on the left-hand side of the screen that is here to help you get started with Rasa. Rasa Copilot comes with specialized knowledge of CALM and provides you will helpful tips, code suggestions, and debugging help as you build. Copilot has the context of your project so you can ask it specific questions or for code snippets that will help you build new skills

This opens a panel that provides real-time visibility into your agent's decision-making as you ask them questions. While you are chatting with your agent, the inspector shows exactly which flows it's executing and how it's progressing through the conversation. You see every decision as it happens, no black boxes.

Yes. When you're ready to build locally or move to production, log in to the playground and continue development with the full Rasa platform. Your playground agent becomes your local development starting point.

Yes. The playground is free to use, no credit card, no trial period. It's designed to help developers explore Rasa quickly and build proof-of-concept agents. All projects built in the playground run on your own Rasa Developer license (see terms here), which not only lets you build with Rasa but also deploy it for small-scale conversational use cases for free.

The playground is your fast entry point—experiment, learn, prototype in minutes. The Rasa Platform is the full collaboration platform for building, deploying, and scaling production agents with team tools, analytics, and enterprise features. Start in the playground, graduate to Pro when you're ready to start working together and shipping. If you are interested in a demo of our full platform, talk to our team.

Rasa utilizes CALM (Conversational AI with Language Models), a dialogue system that combines the flexibility of language models with explicit business logic. The LLM handles natural language understanding and conversation dynamics, while your flows ensure the agent follows your business rules exactly.

CALM (Conversational AI with Language Models) is Rasa's framework that interprets user input, manages conversation flow, and keeps interactions on track. It combines the natural language understanding of LLMs with the reliability and control of predefined business logic.

Rasa provides you with LLM fluency (understanding varied user input and handling natural conversations) and explicit control (business logic, data collection, compliance, and security) in one system. You're not choosing between "smart but unpredictable" or "reliable but rigid."

Flows describe the logical steps your AI agent uses to complete tasks—what information you need from users, when to call APIs, and branching logic based on conditions. Instead of mapping every possible conversation turn, you define key steps. The LLM handles natural language; your flows ensure business rules are followed. You can also define agentic loops within your flows if you want to support more exploratory, prompt-driven conversations.

You define the task steps and business logic in flows. When a user interacts with your agent, Rasa interprets their full conversation context, determines their current position in your flow, and executes the appropriate next step. The LLM provides understanding; your flows provide structure.

Dialogue understanding is how Rasa interprets user messages. For every message, it uses a language model to understand the user's meaning within the full context of the conversation, rather than just as an isolated statement. It generates internal commands that route the user to the correct flow. Because of the language model in play, Rasa can easily understand and handle double-barreled requests, extremely long inputs, back-to-back inputs, and inputs that only make sense in the context of what the user said previously.

No. Rasa uses dialogue understanding (DU), which leverages language models to interpret user input in a conversation context. You define flows (your business logic), and the language model handles understanding user intent without training custom NLU models. Nevertheless, if you still want to use NLU, you can. Rasa's framework still supports NLU classification, and you can even combine it with a language model for a hybrid approach to dialogue understanding.

Yes. Rasa naturally handles interruptions, digressions, clarifications, and topic changes out of the box. In Rasa, we call these conversation patterns—the system flows that handle these common interactions. If a user changes their mind mid-flow or asks an off-topic question, the agent handles it gracefully and can return to the main task when appropriate.

Pure LLM agents can be hard to control and unpredictable. You can't guarantee they'll follow specific business logic, collect required information the same way, maintain compliance, or resist prompt injection. Rasa provides you with LLM fluency for natural conversation, along with explicit control over everything that matters to your business.

Yes. Rasa agents are designed for high-trust, production environments. They provide explainable logic, consistent behavior, and resistance to prompt injection by design. Companies use Rasa in banking, healthcare, telecom, and other domains where reliability isn't optional.

Rasa's architecture separates the interpretation of user input from the execution of actions. Rasa uses a language model to interpret what users are saying and route the user to the right flow(s); however, your flows control what the agent actually does. Users can't bypass your business logic through clever prompting.

With Rasa's playground, we are betting that the learning curve for developers is much gentler. Without much setup, you can start riffing on a templated agent and learn as you go. The copilot explains behavior, the inspector shows you execution flow, and you iterate by testing real changes. No manual-reading required.

Yes. Rasa agents call your APIs, query databases, trigger webhooks, and integrate with third-party services. Flows can include custom actions that execute custom code, making any integration possible. Credential management and out-of-the-box MCP servers are coming soon in the Hello Rasa playground. We don't recommend storing sensitive data or credentials in the playground. If you're ready to add new integrations, download your Hello Rasa agent before adding credentials to your project.

Rasa supports multilingual agents. Since Rasa uses language models for understanding, your agent can handle any language the underlying LLM supports. You don't have to build separate agents; with Rasa, you can provide different answers based on language and locale.

RAG in Rasa combines document retrieval with LLM generation inside the dialogue flow. When a user asks a question that requires external knowledge, the agent retrieves relevant information from connected sources, such as a knowledge base or API, and feeds it to the language model to craft a grounded, context-aware answer.

Rasa allows you to optionally rephrase static (canned) responses within the context of the ongoing conversation. You can enable or disable rephrasing per response, and customize the rephraser prompt to match your brand voice or tone. Use the contextual response rephraser globally to implement a unified brand voice, or customize it selectively for specific answers. Turn it off if you want consistent, identical replies. Overall, it improves conversational fluency and makes interactions feel more natural.

The playground inspector panel shows you real-time flow execution as you chat. You see which flows trigger, what data gets collected, and how decisions are made. Ask the copilot why something happened, and it'll explain the behavior and can offer code to fix it.

Rasa handles unclear input gracefully. Your agent can ask clarifying questions, request more information, or guide the user back to supported flows. You define fallback behavior in your pattern flows, so you can customize what happens when input is ambiguous or off-topic.

In the playground: prompt the copilot to add the feature, paste the generated code, and test it. Locally: edit your flows directly in code. Either way, adding features means defining new steps in flows or extending existing ones—straightforward additions to your business logic.

Start with Rasa's playground to learn by building. For comprehensive documentation, check rasa.com/docs. For community support, join Rasa's community forum. And remember, the playground copilot is also there to answer questions as you build.