Developer Quickstart
Build your first Rasa agent in minutes with the shortest path.
1) Get a free license
2) Create a project and install Rasa Pro
In your terminal:
mkdir rasa-agent
cd rasa-agent
uv venv --python 3.11
source .venv/bin/activate
uv pip install rasa-pro
rasa init --template=basic
Need to install uv? Get it here.
Set your license:
export RASA_LICENSE=YOUR_LICENSE_KEY
This template uses OpenAI as the default LLM provider. To use a different provider, define a model group in endpoints.yml and reference it via model_group in config.yml.
Set your API key as an environment variable:
export OPENAI_API_KEY=YOUR_API_KEY
3) Connect Rasa MCP Tools
Rasa MCP Tools are a set of tools, skills and docs that enable you to interact with your Rasa agent using natural language. They are included in Rasa Pro 3.16 and later.
Run the setup wizard from your project root:
rasa tools init
The wizard creates .rasa/tools.yaml and downloads offline docs and skills for your project. It also generates MCP configuration for your IDE.
Run your IDE from your project root so Rasa Tools can read .rasa/tools.yaml.
Verify your setup
Open your IDE and verify that Rasa Tools appear in your MCP settings. If the wizard configured your client automatically, you should see rasa-tools listed and enabled. If it's not, make sure to switch it on or run it from the command line.
rasa tools run
For manual configuration or more advanced setup details for specific clients (Cursor, VS Code, Claude Code, JetBrains), read more about Rasa MCP Tools here.
4) Build your first agent
Open a chat with your IDE copilot and send this prompt to get started:
Use Rasa MCP tools to inspect this project and briefly summarize what it does in a few sentences.
Then propose 3 **fun, concrete verticals**—think **book recommendation**, **flight booking**, or **concierge**—not “more of the same template.”
For each: one-line pitch, 2 user stories, what you’d add (flows/slots/APIs/sub-agents/MCP servers), effort S/M/L.
Pick the vertical you like best — or suggest your own.