Nodes overview

Node	Type id	Does
URL Fetcher	urlFetcher	Reads pages: one, several, or everything reachable within a depth
Markdown Converter	markdownConverter	Turns pages into clean text
LLM Processor	llmProcessor	Answers a question about the pages, or extracts one record from each
Data Source	dataSource	Stores records (and optionally images)
MongoDB Store	mongoStore	Stores structured records, with optional derived keys
File storage	s3Store	Stores referenced files and attaches blob references
For Each	forEach	Executes an inline graph per upstream record
Combine	combine	Merges branch records by key
Query	query	Reads stored data back
Data Transformer	dataTransformer	Selects or cleans HTML, or extracts JSON values
Table, Chart, Text	table chart text	Show a result

Any node type can be a starting point of a workflow. Option names in the app and API are crawler-core’s, in snake_case; the canvas also accepts camelCase aliases.

Choose the right reference

Each guide explains inputs, outputs, usage, and failure behavior. The generated configuration tables list declared parameter types, defaults, enums, and bounds directly from engine contracts. MCP clients can call list_node_types(type) against their server to discover the deployed schema.