API Reference
Hand-written API reference for every Crux package, organized by package.
The reference is organized by package. Pick the package you're working with — each one has its own landing page with subpath-by-subpath documentation.
Packages
@crux/core
The base package — prompts, contexts, retrieval, indexing, memory, compaction, agents, flows, plans, tasks, skills, scoring, quality, safety, index, lint, runtime bridge, and observability.
@crux/react
Reactive hooks, browser transports, and the SSE server helper for plans, tasks, blackboards, and working memory.
@crux/indexer
TypeScript source intelligence worker for the Project Index.
@crux/local
The local Crux runtime — crux dev, Go services, web devtools hosting, TUI, evals, index, and lint.
Adapters
@crux/ai, @crux/anthropic, @crux/openai, @crux/google — and the adapter interface for custom providers.
Storage
@crux/convex, @crux/upstash — DataStore, VectorStore, and BlobStore adapters.
Plugins
@crux/otel — OpenTelemetry integration for production observability.
@crux/ingest
Node-first ingestion helpers for text, local files, folders, globs, and URLs.
How reference pages are organized
Package landing pages use package names. Pages inside a package use readable API-area names in the sidebar, while the exact import path appears in the page body.
Every reference page follows the same layout so you can scan quickly:
- Overview — what it is and when to reach for it
- Signature — the TypeScript signature of the export
- Parameters — table of options with types and defaults
- Returns — what comes back
- Example — a minimal working snippet
- Hooks emitted — instrumentation events the export fires (when applicable)
- Related — links to the relevant guide and cookbook recipes
If you want narrative how-to content with "when to use" guidance, look in Guides. If you want full working examples, look in Cookbook. The reference is for what does this function do, exactly.