Setting up your Agent Harness for Building with Rasa

README.md

Create a fresh directory, and create two files in it:

1. pyproject.toml - see file contents below. Edit the name and description to match your project.
2. CLAUDE.md - see file content below.

Make sure you have uv installed, then run.

uv sync

To set up the Rasa MCP tools for your coding agent / IDE, run:

uv run rasa tools init

Then start your coding agent (e.g. claude)

CLAUDE.md

Instructions for working in this project.

This is an agent for an airline customer service use case.

Whenever you implement a new feature or make changes, always:

1. train the model
2. have a few conversations with the trained agent to confirm the changes behave as intended.
3. report back the transcripts to stdout as simple markdown.

pyproject.toml

[project]
name = "my-agent"
version = "0.1.0"
description = "Minimal Rasa Agent"
requires-python = ">=3.11,<3.12"
dependencies = [
    "rasa-pro>=3.16.3",
]

[tool.uv]
package = false