Rasa Tutorial — Part 4: Tools: Let Your Agent Run Python
Write your first Rasa Mantle tool: an async Python function in a skill’s tools.py that the agent calls to look up data, here the addresses of a bank’s branches.
- Rasa
- Rasa Pro
Write your first Rasa Mantle tool: an async Python function in a skill’s tools.py that the agent calls to look up data, here the addresses of a bank’s branches.
Write your first Rasa Mantle skill: one skill.md file, with a description that tells the agent when to use it and plain-language instructions that tell it what to say.
Connect a Rasa Pro Mantle agent to Claude: add your Claude API key to .env, set the Anthropic provider in integrations.yml, then train it and check it in the Inspector.
Why Rasa Pro logs “Unknown model name, using cl100k_base encoding as fallback” for Claude models, what Rasa uses that count for, and why you can leave it.
Why the first rasa train warns that agent.yml is missing a unique value for 'agent.id', and how to give your Rasa Mantle agent a readable id of your own.
Run Rasa’s setup wizard, switch your new Mantle agent to Claude, train it and chat with it in the Rasa Inspector. Part 2 of the Rasa Tutorial for beginners.
Rasa Pro has two engines: CALM, built on flows, and Mantle, built on skills. Rasa now recommends Mantle for new agents. Here’s how they differ, and when CALM still makes sense.
Rasa Pro is Rasa’s current framework for AI agents; Rasa Open Source is the original free one, now in maintenance mode. Here’s how they compare and which one to install for a new agent.
Rasa Pro is a code-first framework for building AI agents, and Mantle is its newest engine. Part 1 explains what both are, what the free Developer Edition allows, and what this series builds.