Core Concepts
How ToolCairn thinks
ToolCairn isn't just a search box. These pages explain the ideas it's built on — how tools relate, how queries become ranked answers, how health scoring works, and how feedback keeps the catalog honest.
Start here
Tool Graph
30,000+ tools and the relationships between them — companions, alternatives, dependencies, ecosystems.
ReadSearch Pipeline
How a query becomes a ranked answer: keyword and semantic retrieval, context filters, relevance rerank, selection.
ReadHealth Tiers
Why tools live in tiers (active, stable, slowing, at-risk) and what those tiers actually mean for your decisions.
ReadGoing deeper
Guided Discovery
When the candidate set is genuinely close, the engine asks one clarifying question instead of guessing.
ReadKeyword-Sentence Retrieval
The single biggest accuracy lever in get_stack — a precise keyword sentence per layer.
ReadDevice Auth Flow
How MCP clients sign in via browser device codes, and what the resulting token covers.
ReadEdge Decay
Why old, unconfirmed relationships fade gracefully instead of getting stuck at full strength.
ReadFeedback Loop
Every report_outcome call nudges the catalog. The graph improves with use; bad pairings attenuate.
ReadVersions & Compatibility
How declared peer ranges across registries answer "does X@a work with Y@b?" with a real signal.
Read