Tool: Context.dev — Zelaxy Docs

Context.dev provides web intelligence APIs that return clean, LLM-ready content. Use it to search the web with natural language, scrape any URL to markdown, or crawl an entire site — all returning structured output ready for an Agent to reason over.

Overview

Property	Value
Type	`context_dev`
Category	Tool — Web Scraping
Auth	API Key (Bearer token)

Operations

Operation	Tool ID	Description
Search	`context_dev_search`	Search the web with a natural-language query and optionally scrape each result page to markdown
Scrape Markdown	`context_dev_scrape_markdown`	Scrape a single URL and return clean, LLM-ready markdown content
Crawl	`context_dev_crawl`	Crawl an entire website starting from a URL and return each discovered page as clean markdown

Configuration

Setting	Type	Required	Description
`apiKey`	string (secret)	Yes	Context.dev API key. Passed as a Bearer token. Store as a secret and reference with `{{CONTEXT_DEV_API_KEY}}`.
`query`	string	Yes (Search only)	Natural-language search query, e.g. `"latest AI research 2025"`.
`markdownEnabled`	boolean	No	When `true`, each search result page is scraped and returned as markdown. Defaults to `false`. Applies to Search only.
`url`	string	Yes (Scrape / Crawl)	The full URL to scrape or crawl. Must include `http://` or `https://`.
`useMainContentOnly`	boolean	No	When `true`, strips headers, footers, and navigation from the scraped output. Applies to Scrape Markdown only.
`includeLinks`	boolean	No	When `true`, hyperlinks are preserved in the markdown output. Defaults to `true`. Applies to Scrape Markdown only.
`maxPages`	number	No	Maximum number of pages to crawl (1–500). Defaults to `100`. Applies to Crawl only.
`maxDepth`	number	No	Maximum link depth from the starting URL. Set to `0` to crawl only the start page. Applies to Crawl only.

Outputs

Field	Type	Description
`data`	`json`	Search / Crawl: array of result objects (each containing page content and metadata). Scrape Markdown: object with `markdown` (string) and `url` (string) fields.
`metadata`	`json`	Credit accounting object with `creditsConsumed` (number) and `creditsRemaining` (number) fields.

Example

[Starter] → [Context.dev: Search] → [Agent: summarize findings] → [Response]

Connect a Starter block that accepts a user question, then pass {{starter.input}} as the query to the Context.dev Search block authenticated with {{CONTEXT_DEV_API_KEY}}. Enable markdownEnabled if you want full page content alongside snippets. The Agent block receives {{contextDev.data}} — an array of web results — and synthesizes a grounded answer from real-time sources.

Tips

Choose the right operation — Search is best for open-ended questions where you do not know the exact URL. Scrape Markdown is best when you already have a URL and need its content. Crawl is best for ingesting an entire docs site or product website.
Credit awareness — every operation consumes API credits. Monitor {{contextDev.metadata.creditsRemaining}} in downstream blocks or logs to avoid unexpected quota exhaustion.
Main content only — for Scrape Markdown, set useMainContentOnly to true when passing content to an Agent; it removes nav and footer noise that wastes tokens.
Crawl depth vs breadth — use maxDepth: 0 or maxDepth: 1 with a high maxPages to breadth-crawl a large site; use a deeper maxDepth with a low maxPages to follow a single content thread deeply.

Context.dev