Context.dev

Crawl the relevant pages of a website and return data shaped by the supplied JSON Schema and instructions. Use this when the requested output has defined fields, such as pricing, locations, jobs, or company facts. Use web-scrape-markdown when readable page content is more useful than structured JSON.

web-extract

Find image assets on one web page, including responsive sources, CSS backgrounds, inline SVGs, video posters, and available metadata. Optional browser actions can interact with third-party pages and may cause external side effects, so use them only with explicit user direction. Use this for asset discovery, brand libraries, catalog work, or visual audits; use web-screenshot when a rendered picture of the whole page is needed.

web-scrape-images

Retrieve one detected change with full text diffs, added or removed URLs, semantic evidence, confidence, importance, and the current snapshot when available. Use after either change-listing tool when the complete evidence is needed.

get-change

Page through the successful and failed URL results of a completed batch as JSON. Use the returned cursor to continue when more results exist. For progress before completion, use get-batch instead.

get-batch-results

Retrieve one batch's status, progress, timing, credit accounting, errors, and download links when complete. Poll this after submit-batch; use get-batch-results only after the batch has completed.

get-batch

Retrieve machine-readable company and brand intelligence—logos, colors, descriptions, socials, links, industry, location, and more—from one domain, company name, work email, stock ticker, ISIN, transaction descriptor, or direct URL. Use get-brand for a visual card from a domain; use this tool for raw structured data or non-domain lookups.

brand-retrieve-unified

Retrieve one monitor's configuration, status, schedule, target, change-detection rules, webhook settings, and current baseline by ID. This reads the monitor definition; use list-monitor-runs for execution history or list-monitor-changes for detected changes.

get-monitor

Report credits charged by monitor over a time window, ordered by the monitors using the most credits. Use this for spend analysis and optimization, not for run or change details.

list-monitor-credit-usage

Identify the font families used by a website, with fallbacks and usage statistics. Use this when typography must be matched or audited; use web-styleguide for the broader visual system.

web-fonts

List detected changes across all monitors in the authenticated account, with monitor, type, time, and tag filters. Use this for an account-wide change feed; use list-monitor-changes for one monitor.

list-changes

List monitor runs across the authenticated account. Use this for an organization-wide activity feed or operational overview; use list-monitor-runs when only one monitor matters.

list-account-runs

List changes detected by one monitor, with time and tag filters. Use this for a monitor-specific change history, then call get-change when full diffs, evidence, confidence, and importance are needed.

list-monitor-changes

List recurring monitors in the authenticated Context account, with search, status, target-type, tag, and cursor filters. Use this to find a monitor ID before reading, updating, running, or deleting it; this does not return run history or detected changes.

list-monitors

List the execution history of one monitor, including run status, timing, credits, and whether each run detected a change. Use get-monitor-run for the full details of one run.

list-monitor-runs

List asynchronous scraping batches from newest to oldest, with status, search, tag, and cursor filters. Use this to find a batch ID or review account activity; it does not return page-level results.

list-batches

Send PDF, Office, spreadsheet, presentation, image, code, data, or text file bytes to Context and convert them into clean Markdown for agents, analysis, and RAG. Use this when the source is a file rather than a URL. This transformation consumes Context credits but does not alter the original file.

parse-document

Retrieve a monitor run — Fetches one run for a monitor, including lifecycle status, timing, credits charged, and any detected change.

get-monitor-run

Queue an immediate check for a monitor outside its regular schedule. This starts an asynchronous network run, may advance the monitor's stored comparison baseline, and can notify its configured external webhook when a change is detected. Use get-monitor-run or list-monitor-runs to follow its outcome.

run-monitor-now

Turn one known URL into clean, LLM-ready Markdown. This is the default tool for reading or analyzing a page and supports main-content filtering, selectors, links, images, PDF page ranges, dynamic-page waits, and optional browser actions. Browser actions can interact with third-party pages and may cause external side effects, so use them only with explicit user direction. Use web-search when the URL is unknown, web-crawl for several linked pages, or submit-batch for large asynchronous jobs.

web-scrape-markdown

Fetch the raw HTML of one known URL. Use this only when DOM structure, attributes, scripts, or markup are required; prefer web-scrape-markdown when the goal is to read, summarize, or answer questions from a page. Supports selectors, dynamic-page waits, and optional browser actions. Browser actions can interact with third-party pages and may cause external side effects, so use them only with explicit user direction.

web-scrape-html

Search live and historical news for one company, identified by exactly one name, domain, ticker with an optional exchange, or ISIN. Filter by publisher domain or country, article language or type, and publication date; paginate results with stable story IDs and verified entity relevance. Use web-search for broader topics or news that is not tied to one company.

get-news-search

Search the live web and optionally scrape ranked results to Markdown in the same call. Use this to discover sources, answer current questions, or research a topic when the exact page URL is unknown. When a URL is already known, use web-scrape-markdown instead.

web-search

Retrieve live brand intelligence for a domain and render a visual Context card with its logo, colors, slogan, description, social profiles, industry, location, and key links. Use this when a user wants to see or inspect a company's brand. Use brand-retrieve-unified instead for raw structured data or lookup by name, email, ticker, ISIN, transaction descriptor, or direct URL.

get-brand

Start an asynchronous job to scrape up to 25,000 supplied URLs or crawl a large website as Markdown or HTML. Use this when a synchronous scrape or crawl would be too large. When webhookUrl is supplied, Context sends a completion request to that external endpoint. The call returns a batch ID, not page results; use get-batch until it settles, then get-batch-results. Supply an Idempotency-Key when a submission may be retried.

submit-batch

Update an existing monitor's configuration, including its optional external webhook. Changes affect future runs, and changing the target or change-detection definition creates a new baseline. Use get-monitor first when the current configuration must be preserved.

update-monitor