The Tooling

How to read thisA router, not a reading list

This is a Reference page — you consult it, you don't work through it. Every element of an AI system has a small set of go-to tools; for each element you get the tools that matter and why you'd reach for them. The rule of thumb — use a managed API to prove quality, a framework to move fast, and only build custom when a real constraint (cost, latency, control) forces it.

The tools are not the point. Every tool name is a link to the thing it does, because "which vector database" is a question you can only answer once you know what a vector database is for. Three states, and they look different on purpose:

• Name with an arrow (→)opens the Foundry page that teaches the concept behind that tool. Pinecone lands on Vector databases, Lakera on Prompt-injection defense, LiteLLM on AI Costing. You are still on this site.
• Name with a corner arrow (↗)leaves for the vendor's official docs, in a new tab. Used only where Foundry genuinely has no page on it.
• Grey, dashed, no arrowno destination. It is not clickable and does not pretend to be.

Each card also carries a Learn it here row: the two or three pages that teach the element itself, regardless of which tool you pick. That is the row to follow if you came to this page and realised you were shopping before you understood the problem.

Right now: 43 of 54 tools open a Foundry deep-dive, 11 go to the vendor's own docs because nothing here teaches them yet, and 0 are inert. These numbers are counted from the data at build time, not typed in.

The stack, tool by toolEvery element, in nav order

Cards run in the order the top nav runs — Foundations, Models, Ground, Build, Operate — and each card is coloured by the group that owns it. An element's colour here is the same colour it has in the menu and on its own page, so you can find where a tool belongs without reading a word.