Griptape Trade School is the education arm of Griptape, a Python framework for building generative AI applications with abstractions for LLMs, RAG, agents, pipelines and workflows. The courses are free and there are six or so: building a chatbot with rulesets, comparing movies using workflows, image generation with pipelines, integrating with ShotGrid, an image query exercise, and one on structures calling structures. The pedagogy is the strong part and I want to give it proper credit, because it is done better here than in most vendor material. Every course is project-based, meaning you apply concepts by building an actual thing rather than reading about abstractions and hoping they cohere later.
More specifically, the courses display real code with line numbers and highlight exactly what changed at each step. That sounds trivial. It is not. The single most common failure in programming tutorials is the reader knowing what to type and not knowing where it goes, and showing the diff rather than the final file fixes it.
Whoever designed this understood what actually confuses learners. The ShotGrid course is worth singling out. ShotGrid is production management software for film and VFX, and a course about building a Griptape tool that turns chatbot commands into automated pipeline tasks is refreshingly specific. Most vendor education stops at a weather API demo.
Something that engages with a real industry workflow is more instructive precisely because it is less generic, and if you work in that pipeline it is immediately applicable. Now the difficulty, and it is the one that determines the rating. Griptape has not won its market. The Python agent framework space consolidated hard around LangChain and LlamaIndex, with CrewAI, AutoGen, Pydantic AI and others taking meaningful positions, and Griptape's search interest has been flat to declining for a year on the data I looked at.
That is not a comment on code quality, which is fine. It is a comment on ecosystem, and ecosystem is most of what a framework gives you. Integrations, Stack Overflow answers, blog posts about your specific error, people who already know it when you hire them. Choosing a framework outside the top few means solving problems alone that others solved for you.
There is also drift. Some course material references abstractions such as ToolkitTasks that have since evolved in the framework, which is the ordinary fate of vendor tutorials that are not continuously maintained and is still friction you will hit. And the transfer problem applies as it does to any single-framework course. You learn Griptape's particular decomposition of agents, pipelines and workflows, and how much of that is a general truth about agent systems versus one team's API design is not made clear.
My honest recommendation. If you are picking a Python agent framework for new work in 2026, this is not where I would start. If you want to spend an hour seeing what genuinely good project-based technical teaching looks like, or you work in VFX and the ShotGrid angle is directly relevant, it is free and it is well-made. Three point four.
Better teaching than the framework's current position deserves.