AI Engines

Frameworks · runtimes · agent platforms

AI Engines are the frameworks and runtimes that power AI features — the layer that owns agents, tools, workflows, memory, and model calls. This section of Knowledge holds a handbook per engine: deep reference you read to understand and build, not a step-by-step procedure.

Note

Engines vs. the AI System. The AI System topic documents our agent setup — the hooks, skills, MCP servers, and rules that drive Cursor/Claude. AI Engines documents the third-party frameworks an app is built on. Different layer, different audience: AI System = how the assistant works; AI Engines = what your product runs on.

Handbooks

• Mastra — the TypeScript agent framework (agents, tools, workflows, memory, RAG, eval, observability) built on top of the Vercel AI SDK. Start with the overview, then concepts → architecture → building agents → workflows → deployment.

More engines will be added here. Each handbook lives in its own folder under src/content/docs/ai-engines/.

How a handbook is laid out

Every engine handbook follows the same spine — read it top to bottom the first time, or jump straight to the page that answers your question:

1. 1

   Overview

   What it is, who it’s for, when to reach for it, and the mental model.

2. 2

   Concepts

   Each core primitive defined crisply — the vocabulary you’ll reuse everywhere.

3. 3

   Architecture

   How the pieces fit and how a request flows end to end.

4. 4

   Building agents

   The minimal end-to-end path to a working agent.

5. 5

   Workflows

   Multi-step orchestration, branching, and human-in-the-loop.

6. 6

   Deployment

   Dev → build → ship, deployer targets, and production concerns.

The same six-page spine for every engine, so you always know where to look.