Changelog

​

Access policies

• Control which websites your API requests can reach. An access policy is an allowlist or a blocklist of domains. Each domain includes its subdomains, and *.gov covers a whole extension.
• List policy IDs in the new policies param on Scrape, Crawl, Map, Search, Answers, News, Brand, Styleguide, Fonts, Prefetch, batches, and monitors. Or set applies_to to apply a policy automatically to chosen API keys or your whole account.
• A blocked request returns 403 ACCESS_POLICY_DENIED and isn’t charged. Search, Answers, and News remove blocked results instead. Redirects and embedded frames are checked too.
• Manage policies with the free access policy endpoints, which use the new access_policies:read and access_policies:write key permissions. Access policies are enabled per organization; contact us to turn them on.

​

Ultra mode for People Search

• Set mode: "ultra" on People Search for detailed requests that combine several criteria. Ultra plans a deeper search, judges each candidate against every criterion in your query, and returns a match.confidence score from 0 to 1 with each person.
• Ultra accepts queries up to 5,000 characters and costs 50 credits per page with results. fast stays the default at 10 credits. Complex ultra searches can take up to 30 seconds.
• Try both modes in the playground.

​

Page numbers in PDF Markdown

• Set includeNumberedPages to start each PDF page in the Markdown with a <!-- page N --> line that uses the page’s original number, so you can cite pages or split a document by page.
• Use includeNumberedPages=true on Parse, or sharedParams.parsers.pdf.includeNumberedPages on Scrape and batches. Scrape responses with page markers are never served from cache.
• The Parse and Scrape playgrounds have a toggle for it.

​

Exact phrases in People, Company, and News search

• Put a phrase in double quotes to require it word for word in People Search, Company Search, and natural-language News Search.
• Join quoted phrases with OR to accept either one, as in "payroll" OR "invoicing", and put - before a quoted phrase to exclude it.

​

People Search

• Search people with a natural-language query, such as “VPs of sales at solar companies in California”, and get up to 10 people per page with their current role and employer, location, experience, skills, schools, and profile links.
• Narrow results by country and years of experience, and page through them with a cursor.
• Each page with results costs 10 credits; searches with no matches are free. People Search is available on all plans and in the playground.

​

New self-serve plans

​

Company search

• Search companies with a natural-language query and get each company’s identity, headquarters, and headcount, with pagination.
• Each request costs 2 credits; requests with no results are free.

​

Research papers in Search

• Search now includes indexed research papers for research-style queries. Papers are skipped when domain filters are set.

​

Longer News descriptions

• Natural-language News results now return extended descriptions, combining the publisher’s summary with snippets that matched your query, up to 2,000 characters.

​

Choose where Markdown comes from

• Many docs sites publish a Markdown version of each page. Set markdownParams.sourcePreference on Scrape and batches, or sourcePreference on Crawl, to rank html-conversion and markdown-direct. The first usable source wins.
• The default, ["html-conversion", "markdown-direct"], keeps the current behavior. Use ["markdown-direct", "html-conversion"] to prefer the site’s own Markdown, which often keeps code blocks and heading anchors cleaner.
• With only markdown-direct, a page that publishes no Markdown fails with NO_PUBLISHED_MARKDOWN and is not billed. Link and image options require html-conversion.
• Crawl now also falls back to published Markdown when an HTML conversion is nearly empty, as Scrape and batches already did.

​

Follow only chosen languages in batch crawls

• Set controls.languages on a start-URL batch crawl to same_as_start_url or a list such as ["en"], so translated copies of a site don’t use up maxUrls.
• Links whose language can’t be determined are still followed.

​

No per-minute limit on Free

• New Free organizations no longer have a per-minute limit; the 2 concurrent requests limit still applies. Existing and custom allowances are unchanged. See Rate limits.

​

Fewer rate-limit digests

• The daily rate-limit digest is now sent only when more than 20% of your organization’s requests in the past 24 hours were rate-limited, up from 5%.

​

Exclude URLs from crawls and site map discovery

• Skip URLs with an exclude pattern:
  • excludeUrlRegex on Crawl and Map URLs
  • controls.excludeRegex on batch crawls from a start URL or sitemap.
• A URL that matches the exclude pattern is skipped even when it matches the include pattern, so you can drop sections such as /blog/. The start URL is always scraped.

​

Natural-language News search

• Ask News in plain words, like “chip export restrictions to China”, with searchBy: { type: "natural_language", query } and get up to 100 matching articles ranked by relevance.
• Every News filter still applies, and the playground now opens on natural-language search.

​

Credit exhaustion returns 402

• Authenticated requests rejected for insufficient credits now return 402 USAGE_EXCEEDED, including batch reservation failures. Previously these failures returned 401 USAGE_EXCEEDED. The error code and envelope fields are unchanged; clients that checked for 401 should handle 402. The public API stays on /v1.
• Additive error_data reports known credit facts, retry conditions, and available refill information. Rejected operations remain uncharged; unknown fields are omitted.
• MCP billing failures return model-visible tool errors in text and structured content without OAuth challenges. Usage inspection remains free at zero balance. Recovery links open subscription settings; paid-feature entitlements and temporary throttling remain distinct.
• Concurrency rejections retain 429 RATE_LIMITED and add structured error_data with the effective limit, remaining slots, retry delay, zero charge, and recovery by waiting for capacity or increasing the limit. REST and MCP use https://context.dev/dashboard/subscription for account settings. Per-minute throttling is unchanged.

​

TikTok parsers

• Pull TikTok profiles, videos, feeds, comments, followers, and user search results with seven new TikTok parsers.
• Use them through Scrape Anything or the playground, without a TikTok login. Video results include a video_url you can download directly.

​

People Enrich on every plan

• People Enrich now works on every plan, including Free, so you can match a person from identity clues without upgrading.
• Pricing stays at 20 credits per match.

​

Instagram parsers

• Extract public profiles, posts, reels, and hashtag, keyword, or location feeds with the three new Instagram parsers.
• Use them through Scrape Anything or the playground, without an Instagram login.

​

Google Maps reviews

• Extract up to 1,000 reviews for a place with the Google Maps reviews parser, in the API or playground.
• Choose newest, oldest, most relevant, highest-rated, or lowest-rated order; the default is 100 newest reviews.

​

G2 reviews

• Extract the 25 most recent reviews for a product with the G2 reviews parser, including ratings, reviewer details, likes, dislikes, and problems solved.
• Use it in the API or playground with optional star-rating filters in the URL; pagination is unavailable.

​

Language filters for Search and News

• Set Search languages, such as ["en", "de"], to narrow results by language; the default is "all".
• Search and News language filters support 74 languages; results whose language cannot be detected remain eligible.

​

Brand search filters and richer results

• Filter Brand search by country and primary language, such as country=US&language=english; both filters must match when combined.
• Results now include description, country, and language when available.

​

Daily rate-limit digest

• Rate-limit account emails now arrive as one daily digest that names each limit you hit and exactly how many requests it rejected.
• The digest is sent only when more than 5% of your requests in the past 24 hours were rate-limited, so occasional throttling no longer fills your inbox.

​

Scrape Anything batches up to 100,000 URLs

• Submit up to 100,000 URLs, or crawl a site, in one asynchronous batch.
• Combine Markdown, HTML, screenshots, images, bytes, parsers, product data, highlights, and JSON extraction with the same per-page options as Scrape Anything.
• Create jobs and inspect results from the dashboard too.

​

List browser profiles

• GET /org/browser/profiles lists your organization’s browser profiles with their creation time, newest first. It costs no credits and needs logs:read.
• Add an optional description when you create a profile to label it in that list.

​

Updated request limits

• New Free organizations receive 2 concurrent requests and 10 requests per minute. Both limits apply across the organization.
• Existing allowances are preserved through the rollout and routine renewals. Plan changes apply the new plan’s base limits; explicit overrides remain until cleared.
• The dashboard shows your effective request limits.

​

Eight prebuilt parsers

• Prebuilt parsers extract Google Maps businesses, Google Play and App Store app details and reviews, emails, phone numbers, and links without custom extraction rules.
• Set formats.parse: true, parseParams.type: "prebuilt", and a parser name, or choose a parser in the playground.

​

Request fewer Search results

• Search accepts numResults from 1–100, down from a minimum of 10. The default remains 10.

​

Authenticated scraping with browser profiles

• Sign in once and reuse a saved browser session with sharedParams.profile on Scrape Anything. Browser profiles are available on paid plans and belong to your organization.
• Create and reuse profiles in the playground. Profile requests bypass shared caches and do not support zero data retention.

​

Search understands relative time

• When you omit freshness, Search can infer a publication window from queries such as “today” or “this week”.
• An explicit freshness setting takes precedence.

​

Multilingual Map search

• Map search matches search terms across website languages to find relevant pages even when their content uses a different language from your query.

​

Longer batch result retention

​

Monthly credits for every Free organization

• Free organizations receive 1,000 credits monthly, whether they signed up with a work or personal email.

​

Search description length

• Set descriptionMaxCharacters on Search to cap each result’s description.
• When it is above 0 and the query has no domain or freshness filter, news-index results fill that space with the article passages that best match your query.

​

Search page reads

• Search page reads now default to a 20-second deadline per page, plus any waitForMs, instead of 45 seconds. Set markdownOptions.timeoutOpts to change it.
• Pages that serve an anti-bot challenge report WEBSITE_BLOCKED on markdown.code and highlights.code.

​

JSON extraction on long pages

• JSON extraction now also reads head metadata (description, canonical URL, language, author, keywords, dates), counts JSON-LD and meta tags separately from the Markdown limit, and keeps the sections that best match your schema on long pages.
• A successful extraction from a page that did not fit includes json.isTruncated: true.
• Numbers that do not appear on the page and punctuation-only strings are dropped from optional or nullable fields.

​

Answers sources and partial answers

• sources lists the pages and company or profile records read for the answer, followed by URLs cited in json_content.
• With return-partial, a model error after evidence was gathered returns a partial answer instead of failing.

​

Parse and credit usage

• Parse returns 400 INPUT_VALIDATION_ERROR instead of WEBSITE_ACCESS_ERROR for an upload it cannot read or that has no extractable content.
• GET /org/usage and GET /org/usage/history return your credit balance and usage over time. They share the request logs rate limit.

​

Product image deduplication

• Set productParams.dedupeImages: true to drop visually duplicate product images, keeping the largest copy.
• Product adds 1 credit, with no other product surcharge.

​

Removed EXTERNAL_PROVIDER_ERROR

• EXTERNAL_PROVIDER_ERROR is no longer returned. Upstream failures return 500 INTERNAL_ERROR (Answers and People previously returned 502), failed Scrape json outputs report INTERNAL_ERROR, and Answers time limits return 504 REQUEST_TIMEOUT.

​

Longer Search cache windows

• Search Markdown now defaults maxAgeMs to 180 days (15552000000 ms), or 365 days when only highlights are requested.

​

Highlights on Search

• Search returns passages relevant to your query with highlightsOptions.enabled. Each result reports highlights.code, like markdown.code.
• Search Markdown and highlights each add 1 credit per 10 results when at least one result returns them.
• Search Markdown now defaults maxAgeMs to 15 days (1296000000 ms).
• Scrape highlights now add 1 credit instead of 3.

​

Concurrency limits

​

Browser actions on page monitors

​

Longer scrape cache windows

​

Answers page caching

• Answers can reuse cached page reads: up to 30 days old in fast and 7 days in ultra.

​

Plan changes and account limits

• Upgrades take effect immediately and preserve your remaining credits. Downgrades take effect at renewal and can be cancelled beforehand.
• Review billing for plan changes, Auto-Topup, and spend alerts.
• Rate-limit reminder emails help you identify sustained throttling.
• Free organizations receive an email when repeated errors disable their keys. Fix failing requests and contact support before resuming traffic.

​

Agent Feedback

• Agents can report problems with API calls, documentation, SDKs, or the CLI.
• Include a category, a short note, and the affected request ID or URL.
• Submit feedback with any API key, at no credit cost.

​

Auto-Topup in smaller blocks

​

Highlights format on Scrape

• Find the passages most relevant to a question or topic with formats.highlights and highlightsParams.query.
• Set maxCharacters to limit the total length. Highlights also work with zero data retention.
• Adds 3 credits only when passages are returned.

​

Extract product data with scrape

• Extract product details with formats.product, including prices, images, and variants.
• Results identify whether the URL is a product page and include a normalized product record when available.
• Adds 1 credit.

​

Scrape Anything

• Scrape a page once for Markdown, HTML, screenshots, images, original bytes, or selected fields. Choose any combination with formats.
• Control freshness with maxAgeMs; use 0 to fetch fresh content.
• Costs 1 credit, or 2 with browser actions, plus 1 per PDF page recovered by OCR.

​

Map website URLs

• Discover a site’s URLs with available titles, descriptions, keywords, and languages.
• Filter and search the list, include subdomains, or set a maximum number of links.
• Costs 1 credit, or 2 when using search.

​

SDK updates

• All five SDKs now support Scrape and Map URLs.
• Replace the old per-format scraping methods with the unified Scrape and Map methods.
• Existing HTTP routes keep working. The CLI does not yet include these two commands.

​

Bytes with rendering and cache controls

• Download rendered HTML bytes after a waitForMs delay. Other resources keep their original bytes.
• Downloads are cached for one day by default, up to 30 days. Use maxAgeMs=0 to fetch fresh content.
• Requests with credentials or zero data retention bypass caching.

​

Screenshots with custom headers

• Send custom target headers, including authorized cookies, when taking screenshots.
• Requests with custom headers bypass caching and return an inline PNG data URL.

​

Country routing for page images

• Use country to extract page images from a selected region.
• The country applies to both the page and its image downloads.

​

SOC 2 Type II compliance

• Context.dev is now SOC 2 Type II compliant.
• Review security policies, subprocessors, and access to compliance evidence in the Trust Center.

​

Screenshots from a full URL

• Capture a page from its full URL with GET /web/scrape/screenshot, for 1 credit.
• See the screenshot guide for capture options and examples.

​

Extract fields with CSS

• Use CSS selectors to return structured fields alongside a page’s HTML, with no additional credits.
• extractRules supports text, attributes, lists, and nested objects.

​

Zero data retention for AI endpoints

• Zero data retention now supports Answers, Search, People Enrich, Images, and Styleguide.
• Enable it for your organization, send zdr: "enabled", and verify X-Context-ZDR: true on supported requests.

​

API keys during onboarding

• Onboarding now picks up newly created API keys without getting stuck waiting.
• If loading fails, you can retry; setup prompts become copyable once the key is ready.

​

More reliable scraping and API errors

• Pages with large response headers can now be scraped.
• Temporary gateway failures return a JSON 503 response. See troubleshooting for recovery guidance.

​

More reliable payments

• Fixed duplicate Auto-Topup charges when several requests reach a low balance together.
• Plan upgrades also wait for successful payment, with a clear recovery path when payment fails.

​

Startup and nonprofit program

• Eligible startups and registered nonprofits can apply for 30% off their first year on a self-serve annual paid plan, with a private Slack Connect channel.
• Approval is required; overages are excluded, and existing agreements stay unchanged.

​

Reinforcement learning grants

• Teams building reinforcement learning systems can now apply for Context.dev data credits.
• The program page explains eligibility and how to submit a research proposal.

​

Timeout controls and partial results

• Set a request deadline and choose whether to fail or accept partial results with timeoutOpts.
• Partial results are marked with partial: true or finalDOMState: "still-loading". Bytes and Prefetch support fail behavior only.
• September 16 update: partial HTML, Markdown, Images, Screenshot, and Styleguide captures accept budgets from 5,000 ms.

​

Download raw resource bytes

• Download images, PDFs, HTML, and other resources as base64, with file size, content type, final URL, and origin status.
• Supports files up to 20 MiB, custom headers, country targeting, and zero data retention.
• Costs 1 credit per download.

​

Parse documents up to 50 MiB

• Parse now accepts uploads up to 50 MiB, twice the previous limit.
• Conversion remains 1 credit, plus 1 credit per PDF page recovered by optional OCR.

​

Request logs and search results

• Request logs redact recognized credentials in request fields, headers, and URLs.
• Web Search results always include a description string, empty when no snippet is available.

​

Advanced monitor controls

• Focus page monitors on specific content with include and exclude selectors.
• Changing filters starts a new baseline. Empty filtered results fail the run and preserve the previous baseline.
• The dashboard adds controls for whitespace, webhook retries, and extraction depth.

​

Answers: web research in two modes

• Ask a question about the live web and receive JSON with source URLs.
• Fast: 10 credits with a 30-second budget. Ultra: 100 credits with a 50-second budget.
• The API defaults to Ultra; the playground defaults to Fast.

​

Restricted API keys and per-key usage

• Restricted API keys let you grant only the data, log, monitor, and batch access an integration needs.
• Existing keys keep full access.
• Search and filter keys, then open their usage charts and request logs from the dashboard.

​

Team roles and deactivation

• Manage Owner, Admin, and Member roles, transfer ownership, and deactivate or reactivate teammates.
• Deactivation revokes the person’s credentials and prevents invitations or domain auto-join from restoring access.

​

Formatted Slack webhook notifications

• Send monitor and batch events directly to Slack or GovSlack for readable messages with resource links.
• Other webhook destinations continue receiving JSON.

​

More accurate freshness controls

• Brand retrieval and Styleguide now honor maxAgeMs: 0 and values under one day.
• Set 0 to fetch fresh data.
• The three-month default and one-year maximum are unchanged.
• See Cache freshness.

​

Updated Free credit allowances

• Free organizations with a work email now receive 1,000 credits monthly.
• Personal-email organizations receive 250 credits once, without monthly refills.
• This replaces the September 6 allowance.
• Superseded October 6: see current Free credits.

​

Request log endpoints

• Find requests by time, endpoint, key, status, error, or tags, and retrieve an individual call.
• Responses now include request_id and X-Request-Id for tracing.
• Log endpoints cost 0 credits.

​

Better blocked-page detection

• Scraping now recognizes more CAPTCHA, login, and access-denied pages, including pages that return HTTP 200.
• These return WEBSITE_BLOCKED instead of being billed and cached as usable content.
• Brand cache metadata also reflects the age of the underlying scrape more accurately.

​

Context.dev for Gemini CLI

• Connect Gemini CLI to search, scrape, extract, parse, and retrieve brand data from the terminal.
• The official extension uses browser-based OAuth.

​

Context.dev for Composio

• Use Context.dev through Composio for scraping, extraction, parsing, brand data, batches, and monitors.
• Connect an API key to make the toolkit available to your agent framework.

​

Context.dev for Replit

• Connect Context.dev to Replit Agent to search, scrape, and retrieve company data while building your app.

​

Webhook retries, delivery history, and replay

• Configure automatic webhook retries for batches and monitors with webhook.retry.
• Replay failed deliveries for up to seven days, without rerunning the job or spending credits.
• Inspect 30 days of delivery history and attempts in the dashboard.

​

Monthly API credits for new Free accounts

• New Free accounts received monthly allowances of 500 credits with a work email or 250 with a personal email.
• Refills preserved balances above the allowance.
• Superseded September 11: see current Free credits.

​

Higher default monitor limits

• Monitor allowances increase from 2 to 10 on Free, 50 to 100 on Developer, and 500 to 1,000 on Pro.
• Scale’s default is 10,000 monitors.

​

More reliable brand profiles and SVG logos

• Brand retrieval handles redirects, placeholder sites, and error pages more accurately.
• SVG logos now render more reliably outside the original website.

​

Clearer credit counters and dashboard support

• Dashboard counters distinguish monthly allowances from bonus balances, and Support chat is available after onboarding.
• Monthly refills also preserve manually granted balances above the allowance.

​

Zapier 3.1

• Zapier 3.1 adds event triggers and actions across scraping, extraction, batches, monitors, and brand data.
• Start a Zap from a website change, monitor run, or finished batch.
• Existing 3.0 Zaps keep running.

​

Downloadable captured font files

• Styleguide now returns captured font files through stable CDN URLs in fontLinks, so downloads work outside the source website.
• See Extract a design system.

​

More playground languages

• The playground’s language selector now includes all 120 languages supported by the API, plus automatic detection.

​

Better crawls of redirected documentation sites

• Crawl follows sites that redirect to a dedicated documentation subdomain and continues beyond the first page more reliably.

​

Context.dev on Zapier

• Connect Context.dev to your other apps with the Zapier integration.
• The guide includes setup, available actions, and automation recipes.

​

URL mapping can include subdomains

• Set includeSubdomains: true to discover URLs across a domain and its public subdomains.
• The default remains false; the option is also available in the playground.

​

Enterprise SSO on the Scale plan

• Scale organizations can connect an identity provider from Organization Settings.
• Owners create a setup link valid for 24 hours.
• Verified company-domain teammates are routed to the provider at sign-in. Password sign-in remains available.

​

Create additional organizations

• Use the dashboard’s organization switcher to create another organization when every organization you own is on a paid plan.
• Each has separate members, API keys, and billing.

​

Cache metadata on every cacheable response

• Cacheable responses now include cache_metadata with a hit, miss, or zdr status and age_ms, so your application can see how fresh the data is.

​

Context.dev for Claude

• Connect Context.dev to Claude from the Connectors Directory to search, scrape, crawl, extract, parse, and monitor the web.

​

Dashboard dark mode

• Choose light, dark, or system appearance from the dashboard user menu.
• Your preference is remembered across visits.

​

More reliable slow-page scraping

• Slow-loading pages have more time to finish browser rendering, and repeat scrapes can reuse a working access route for sites with stricter protection.

​

Pause your plan for up to three months

• Pause a paid plan for one, two, or three months, starting at renewal.
• Monthly charges, included-credit allocations, and Auto-Topup pause. Existing credits remain usable.
• Cancel a scheduled pause, or pay to resume early and restore monthly credits.

​

Office documents with incorrect content types

• Word, Excel, and PowerPoint files served with incorrect content types are now recognized and parsed instead of being treated as webpage HTML.

​

Timestamped YouTube transcripts

• YouTube transcripts now include timestamps roughly every 30 seconds, making it easier to trace a passage back to the video.

​

Payment alerts and plan status

• Organization owners now receive an email when a subscription payment fails, with a link to update the payment method.
• The dashboard also shows scheduled cancellations.

​

Website names in batch lists

• Batch lists now show each job’s target website, making similar jobs easier to distinguish.

​

Scroll pages before scraping

• Browser actions now support scrolling to load more content before capturing Markdown, HTML, or Images.
• Scroll controls are also available in the playground.

​

Duplicate monitors in one click

• Choose Duplicate from a monitor’s menu to open a new monitor with the same settings.
• Change the target and save.

​

Email verification survives mail scanners

• Verification links now wait for your confirmation, preventing email security scanners from consuming them before you open them.
• This applies to signup and account-email changes.

​

Logo Link wordmarks

• Logo Link now accepts type=wordmark to prefer wide wordmarks over square icons.
• Combine it with theme=light or theme=dark.

​

Richer playground results

• Playground results now show page metadata, freshness, and browser-action outcomes, with clearer previews for scrape, extract, brand, and parse requests.
• Parse displays its OCR credit cost before you run it.

​

Context.dev for ChatGPT

• Add Context.dev to ChatGPT to search, scrape, crawl, extract, parse, and monitor the web from a conversation.

​

Website privacy controls

• Choose which analytics cookies to allow through the website’s consent controls.
• Optional analytics wait for your consent.

​

Context.dev for Cursor

• Install the Cursor plugin to give your coding agent access to Context.dev’s web tools.

​

Context.dev for LangChain

• Use LangChain’s Context.dev integration for search, scraping, extraction, parsing, brand data, monitors, and batches.
• Install the langchain-context package to get started.

​

Clear obstructive popups from screenshots

• Set clearPopups=true to dismiss obstructive banners, popups, and overlays before taking a screenshot.

​

Context.dev for OpenClaw

• Connect OpenClaw for web search and scraping, with optional OAuth access to Context.dev’s public MCP tools.

​

Exclude more domains from web search

• Web Search now accepts up to 100 excluded domains, giving you more control over which sources appear in results.

​

More accurate brand colors, logos, and contacts

• Brand profiles better match website colors, preserve original logos, and filter placeholder contact emails.
• Redirects to shared hosting platforms no longer replace the requested company’s brand with the platform’s.

​

News Search launches in beta

• Find company news by name, domain, ticker, or ISIN.
• Filter by publisher, date, article type, language, or country.
• Successful searches cost 1 credit per 10 returned articles. A successful search with no articles is free.

​

More control over Brand Search matching

• Brand Search adds autocomplete, queryBy, and typoTolerance.
• Choose name or domain matching, prefix matching, and tolerance for up to two typos.
• These controls are also in the playground.

​

Change your account email securely

• Change your email from Personal Settings after confirming your identity and verifying the new address.
• Your previous address receives a notification, and you stay signed in.

​

Context.dev for Eve

• Add Context.dev to Eve with eve add connection/context for search, scraping, extraction, and brand tools.
• Each user authorizes their account in the browser.

​

Screenshots now cost 1 credit

• Screenshot requests now cost 1 credit.
• See Take a screenshot for capture options and examples.

​

Earn extra credits

• Eligible Free accounts with a personal email can earn bonus credits by following Context.dev or submitting a public post for review.
• The dashboard shows the available rewards and submission status.

​

Better text encoding

• Pages, feeds, and sitemaps now respect declared character encodings, reducing garbled text on sites that use older encodings.

​

Prefetch styleguides ahead of time

• Queue a styleguide with type: "styleguide" on Prefetch to reduce the wait for a later Styleguide request.
• Prefetching costs 0 credits.

​

HTML alongside Markdown in one scrape

• Set includeHTML=true to return HTML and Markdown together from one scrape.
• Markdown batches support the same option, also available in the playground.

​

Page headings and main content

• HTML and Markdown responses now include metadata.headings in document order for tables of contents and content routing.
• Main-content filtering also preserves the page’s <main> landmark.

​

People Enrichment API

• Find a person from a social URL, name and company, or work email, with a 0–100 match score.
• The beta is available on paid plans and costs 20 credits per successful match.
• Free and disposable email inputs are rejected without charge.

​

Employee counts on brand data

• Brand profiles now include employees with a headcount range and an exact count when known.
• Unknown values are null.

​

Relevance search for URL mapping

• Give Map URLs a search phrase such as pricing and plans to return matching pages in relevance order.
• Searched requests cost 2 credits instead of 1.

​

Per-page OCR billing and clearer PDF errors

• OCR adds 1 credit per recovered PDF page to the 1-credit request cost.
• Enable OCR with ocr=true for Parse or pdf[ocr]=true for scraping.
• Image-only PDFs return PDF_IMAGES_ONLY when OCR is needed.

​

Clearer timeouts for new brand lookups

• A brand lookup on a new domain with a timeout below 10 seconds now returns 422 COLD_DOMAIN_TIMEOUT_TOO_LOW immediately.
• Timeout responses suggest prefetching the brand before you need it.

​

YouTube pages become structured Markdown

• Scrape a YouTube video or channel to receive its available metadata and description as structured Markdown.
• Videos with retrievable captions also include a transcript.

​

More specific scrape errors

• Identify scrape failures with clearer errors for oversized content, unresolved domains, and intentionally skipped PDFs.
• Oversized embedded files no longer fail the entire page scrape.

​

Faster repeat scrapes

• Repeat scrapes of unchanged pages return faster by reusing the previous Markdown conversion.

​

Brand Search for autocomplete

• Find up to 10 matching companies by name or domain, with logos for autocomplete and lookup forms.
• At launch, requests are free on Pro and Scale and cost 1 credit on Free and Starter.
• August 6 update: name matches rank first, and Brand Search is available in the playground.

​

Context.dev MCP is live

• Connect your AI client over MCP to use Context.dev through browser-based OAuth.
• Setup guides cover Claude, Claude Code, Cursor, Codex, ChatGPT, and VS Code.

​

Usage charts by product

• Dashboard usage charts now separate All, Data APIs, Monitors, and Batches, so you can track activity by product.

​

Monitors recover from failed runs

• Monitors recover from temporary scraping failures instead of pausing after one failed run.
• Resume an already paused monitor by setting its status to active.

​

Credit refills on your billing day

• Monthly credits now refill on your billing day, including monthly allocations on annual plans.
• Your allowance covers a full cycle from subscription rather than resetting on the first of the calendar month.
• Auto-Topup preferences are also preserved across plan changes.

​

Batch scraping for up to 25,000 URLs

• Scrape up to 25,000 URLs asynchronously, using a list, a starting page, or a sitemap.
• Get Markdown or HTML through polling, webhooks, or downloads.
• Credits are reserved for the accepted page budget. Only successful pages are charged; unused credits are released.
• Use item IDs and idempotency keys to match results and retry submissions safely.

​

Connect AI agents over MCP with OAuth

• Connect an AI client over MCP and authorize it in your browser with OAuth.
• The consent screen identifies the client and requested access; no API key needs to be pasted into its configuration.

​

Monitor the changes that matter

• Give page monitors instructions describing which changes matter, such as pricing changes rather than rotating testimonials.
• Semantic detection filters irrelevant changes; empty or blocked snapshots fail the run without replacing a valid baseline.

​

Usage alerts for overages

• Set an Auto-Topup budget and receive email alerts at selected spending thresholds.
• Alerts are a soft cap: API requests, monitors, and Auto-Topup continue after the budget is exceeded.

​

Simpler monitor setup and secret rotation

• Create a monitor with just name and target. Scheduling defaults to daily, with detection selected for the target.
• Check the baseline run using initial_run_id.
• Rotate webhook signing secrets without replacing the monitor. The old secret stops signing new deliveries immediately.

​

Browser actions before scraping

• Click, type, or wait before scraping Markdown, HTML, or Images. Paid plans support up to five ordered actions.
• Standard requests with actions cost 2 credits and bypass caching. Enriched image requests remain 5 credits.

​

Zero data retention

• Eligible scraping and parsing requests now support zero data retention.
• Set zdr=enabled to bypass shared caches and retained content logs, and verify the X-Context-ZDR: true response header.

​

Default organization

• Choose which organization opens by default when you use more than one account workspace.

​

Layout-aware PDF parsing

• Parse PDFs with better tables, reading order, headings, and subscript or superscript text.
• Display equations are preserved as text. OCR can also recover corrupted text layers.
• This update applies to Parse. See the guide for current OCR controls and billing.

​

Faster web dashboard

• Dashboard pages load faster when moving between usage, API keys, team settings, and product views.

​

Request tags for usage tracking

• Tag requests by customer, feature, or environment, then filter charts and logs by those tags.
• Add up to 20 labels without changing pricing or caching. Monitor resource tags remain separate.

​

Parse file bytes to Markdown

• Upload documents directly and get Markdown without hosting a file at a URL.
• Choose format hints, PDF page ranges, and content controls. Conversion costs 1 credit, with optional OCR charges.
• The launch limit is 25 MiB, increased to 50 MiB on September 15.

​

Rate-limit headers on every response

• Authenticated responses now include X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset, so you can pace calls before hitting the limit.
• A 429 also includes Retry-After.
• See Rate limits.

​

Cleaner main-content Markdown

• Main-content filtering removes more navigation, ads, comments, and layout markup from Markdown scrapes.
• The response also includes contentLength, measured in UTF-8 bytes, to help detect empty results.

​

Per-run monitor webhooks

• Subscribe to run.completed for a signed webhook after every completed run, including baseline and no-change runs.
• Existing integrations continue receiving change.detected only.
• Run history adds webhook_deliveries. The singular field remains available but is deprecated.

​

Monitor webhook health tracking

• Receive an email after three consecutive monitor runs with failed webhook deliveries.
• A successful delivery or webhook URL change clears the failure streak. Rate-limited receivers are retried using Retry-After.
• See delivery failures for recovery steps.

​

Join an organization by email domain

• Let teammates with a verified company email domain discover and join your organization without an invitation.
• Manage the setting from the dashboard.

​

Website monitors

• Watch a page, sitemap, or structured data on a schedule and receive signed webhooks when it changes.
• Manage monitors, runs, and changes through the API or SDKs, at no credit cost.
• Successful page and sitemap checks cost 1 credit; extract checks cost 10. Failed or skipped runs are free.

​

Brand lookup from an exact URL

• Retrieve brand data from an exact page with type: "by_direct_url".
• Results use that page’s content only and cost 10 credits on success.
• Domain-lookup options such as maxSpeed, force_language, and maxAgeMs are not supported.

​

One request for brand lookups

• Use one POST /brand/retrieve operation for domain, email, name, ticker, and transaction lookups.
• The type field selects the lookup, giving SDK integrations one consistent entry point.

​

One request for prefetching

• Prefetch now uses one request with a type field for queuing brand data by domain or email.

​

API reference navigation

• The API reference now groups related operations, uses shorter names, and redirects legacy pages to their replacements.

​

Customer referrals

• Share your referral link from the dashboard.
• Referred customers receive 10% off for 12 months, and you earn 10% of their payments for a year.
• The program is available to existing customers without an application and is separate from the affiliate program.

​

Remove duplicate images

• Set dedupe=true on Images to remove near-duplicate images without changing the request price.

​

Scrape PowerPoint presentations

• Scrape PowerPoint URLs into Markdown or HTML, including slide text, tables, and speaker notes.
• Supports PPTX and legacy PPT. Crawls can follow links to presentations.

​

Scrape spreadsheets and CSV files

• Scrape Excel and CSV URLs into tables.
• Workbooks return separate sheet tables, and CSV output preserves an optional title or attribution above the header.
• Markdown, HTML, and Crawl support the new formats.

​

#1 Product of the Day on Product Hunt

• Context.dev reached #1 Product of the Day on Product Hunt.
• Thank you for the feedback and support.

​

Logo Link themes

• Choose theme=light or theme=dark on Logo Link to prefer a matching logo variant.
• Without a theme, the highest-resolution available logo remains the default.

​

Settle animations before capture

• Set settleAnimations=true to stabilize animations before capturing HTML, Markdown, or crawled pages.
• The option defaults to false.

​

Capture light and dark themes

• Set colorScheme to light or dark when taking screenshots or extracting a styleguide to capture the site’s selected color mode.

​

Dashboard refresh

• The refreshed dashboard brings consistent navigation across signup, overview, team management, and API tools, with clearer parameter controls and usage displays.

​

Scrape from a selected country

• Fetch supported scraping, crawling, and screenshot requests from a selected country using a two-letter country code.
• Use it when page content or access varies by location.

​

Capture sections of long pages

• Use scrollOffset to capture a viewport-sized section of a long page starting at a selected vertical offset.
• See Take a screenshot.

​

Annual plans

• Annual billing is now available with two months free compared with monthly billing.
• Choose yearly billing on the pricing page.

​

Account setup for coding agents

• Coding agents can start account setup through auth.md.
• Follow the browser verification step to claim the account and complete setup.

​

PHP SDK

• The official PHP SDK is available through Composer for PHP 8.1 and later.

​

Scrape Word documents

• Markdown, HTML, and Crawl now accept DOCX and legacy DOC URLs, returning structured document content without a separate conversion step.

​

Page metadata

• Scrape and crawl responses now include page metadata such as title, description, language, canonical URL, and structured data.
• Metadata survives content filtering, and relative resources resolve against the final URL after redirects.

​

Hourly usage

• View hourly calls and credit consumption for today and yesterday in the dashboard.

​

Response compression

• Use Accept-Encoding: gzip to reduce transfer sizes for large API responses.
• See Compression for client setup and verification.

​

Identify missing pages

• Scrape and Crawl now return 404 for missing pages, separating them from other website-access failures.
• See Troubleshooting for status handling.

​

Faster usage logs

• Large usage histories load and filter faster, including organizations with millions of request logs.

​

Affiliate program

• The new affiliate program pays 25% of referred customer payments for a year.
• See the program terms for eligibility.

​

Export playground results

• Export a playground response as JSON, copy it, or choose Try in Cursor to open an implementation prompt with the response included.

​

Context.dev CLI

• Use Context.dev from the terminal with the official CLI for macOS, Linux, and Windows.
• Choose JSON or YAML output, transform responses, and pipe results into scripts or coding-agent workflows.

​

Retry guidance for rate limits

• Rate-limited responses include key_metadata with zero credits consumed and a Retry-After header.
• Use the reset delay before sending another request.
• See Rate limits.

​

Credits used and remaining

• JSON responses to recognized API keys now include credits_consumed and credits_remaining in key_metadata, including on validation errors and timeouts.
• Responses to invalid keys omit it.

​

Filter logs by status

• Filter dashboard request logs by HTTP status to find failures and investigate integration issues.

​

Documentation refresh

• Updated guides and navigation make setup and API details easier to find for developers and coding agents.

​

Usage by API key

• View calls and credit usage by API key in the dashboard, with names shown where available.

​

Filter content with CSS selectors

• Use includeSelectors and excludeSelectors to scope HTML, Markdown, and crawl content.
• Exclusion wins when both match; non-HTML documents are unaffected.
• See Content controls.

​

Read underlying XML

• HTML scraping now returns the underlying XML for styled sitemaps and feeds, rather than a browser’s rendered XML viewer.
• The response identifies the detected content type.

​

Custom request headers

• Send custom target headers, including language preferences or authorized credentials.
• Requests with custom headers bypass the shared cache.
• Keep your Context.dev API key separate from headers sent to the target site.

​

Backed by Y Combinator

• Context.dev is now backed by Y Combinator.

​

Cache and search controls

• The Styleguide playground adds cache-age presets.
• Web Search also supports more Markdown, PDF, freshness, and timing controls, with updated SDK examples.

​

Per-minute rate limits

• API limits now use a per-minute window and appear in the dashboard.
• The launch limits are 10 requests/minute for Free, 120 for Starter, 300 for Pro, and 1,200 for Scale.
• See Rate limits for current allowances.

​

Web search

• Search the web and optionally retrieve each result’s Markdown in one request.
• Filter domains and freshness, expand a query, and choose content controls.

​

SOC 2 Type I compliance

• Context.dev is now SOC 2 Type I compliant.
• Review current compliance status and evidence in the Trust Center.

​

Go SDK

• The official Go SDK is available, with typed requests and responses for Context.dev APIs.

​

Dismiss cookie banners

• Set handleCookiePopup=true to dismiss cookie and consent banners before taking a screenshot.
• The option is off by default.

​

PDF parsing controls

• The pdf object replaces parsePDF with controls for whether to parse a PDF and which pages to read.
• Existing requests remain compatible.
• See PDF handling.

​

Set a crawl time budget

• Give Crawl a soft time budget with stopAfterMs.
• It checks the budget between pages and can return the pages collected so far.
• The launch default is 80 seconds, with values from 10 to 110 seconds.

​

Wait for dynamic content

• Use waitForMs on supported web endpoints to allow dynamic content more time to load before capture.

​

Screenshot dimensions and freshness

• Screenshots now accept custom viewport dimensions and cache-age controls.
• Responses include the image’s width and height for layout without downloading it first.

​

Control brand-data freshness

• Control brand-data freshness with maxAgeMs, defaulting to three months.
• At launch, values were clamped between one day and one year.
• Updated September 12: values under one day, including 0 for fresh retrieval, are now honored.
• See Cache freshness.

​

Image caching and enrichment

• Images adds caching and optional dimensions, hosted copies, MIME types, and asset classification.
• Standard requests remain 1 credit; enabling any enrichment option costs 5 credits per call.
• Use maxAgeMs=0 to request fresh results.

​

Better wordmark detection

• Logo extraction better recognizes wordmarks composed of an image and adjacent brand text.

​

Include iframe content

• Set includeFrames=true to include rendered iframe content in HTML, Markdown, and crawl results.
• The option is off by default.

​

Playground controls and docs

• Playground endpoints now link to their documentation and expose their parameters in the UI.

​

Explore more usage history

• Browse paginated request history, filter every endpoint your organization has used, and view longer date ranges.
• Interactive chart legends let you toggle paths and statuses.

​

Scrape PDF URLs

• HTML, Markdown, and Crawl now process PDF URLs, with parsing controls and skipped-page counts.
• PDF scraping launched with automatic OCR; OCR became opt-in on August 6.
• See Parse documents for current controls and billing.

​

Crawl freshness controls

• Set maxAgeMs on Crawl to control how long cached results can be reused.

​

Font-file URLs

• Styleguide responses add fontLinks, mapping font families and weights to available file URLs.
• The result distinguishes Google Fonts and custom fonts.
• See Extract a design system.

​

Filter mapped URLs

• Filter mapped URLs with urlRegex.
• See Map website URLs for examples.

​

Credit charts and log search

• Track credits over time alongside API calls, see usage by endpoint, and search request logs by input.

​

Multiple API keys

• Create and manage multiple API keys from the dashboard to separate integrations and rotate credentials.

​

More social platforms

• Brand social links now include WhatsApp, Telegram, and Line, bringing supported platforms to 31.

​

Brand website language

• Brand profiles include primary_language, identifying the detected primary language of the website.

​

Website crawling

• Crawl a website from a starting URL and return collected pages as Markdown.
• Limit pages and depth, filter URLs, and choose whether to follow subdomains.
• Successful pages cost 1 credit each; the response includes per-page outcomes and totals.

​

Styleguide update

• Styleguide extraction is faster and adds ready-to-use css fields.
• The prioritize parameter is removed; prioritization is handled automatically.
• See Extract a design system.

​

Keep main page content

• Set useMainContentOnly=true to focus Markdown output on primary page content and remove navigation, sidebars, headers, and footers.

​

Discover background images

• Image scraping now discovers CSS background images in addition to standard image elements.

​

Control URL mapping

• Set maxLinks to control URL mapping from 1 to 100,000 links.
• At this release, the returned urls array remains capped at 500.
• See Map website URLs for current limits.

​

Brand.dev is now Context.dev

• Brand.dev is now Context.dev, reflecting the expansion from brand data into web scraping and extraction.
• Existing integrations remain compatible without code changes.
• Earlier entries use the current product name and documentation links.

​

Better logo detection

• Logo detection filters more decorative icons and other shapes that are not the company’s logo.

​

Company lookup by country

• Provide country_gl when looking up a company by name to distinguish companies with the same name in different countries.

​

Credit-based pricing

• API pricing moves from calls to credits.
• Existing quotas convert at 10 credits per call, so 3,000 calls becomes 30,000 credits.
• Standard brand lookups remain 10 credits, while lighter scraping operations cost 1 credit.

​

Scraping in the playground

• Try HTML, Markdown, Images, and URL mapping directly from the dashboard playground.

​

HTML, Markdown, Images, and URL mapping APIs

• Four new APIs turn URLs into HTML, Markdown, images, or a website URL inventory.
• Scraping includes automatic proxy handling without a separate proxy charge.

​

Verify transaction matches

• Set high_confidence_only=true on transaction lookups to request extra verification of the merchant match.
• It defaults to false.

​

Styleguides from an exact page

• Extract a styleguide from an exact page with directUrl.
• Provide either a domain or a direct URL.

​

Lower API latency

• API requests return faster across endpoints.

​

Larger plan quotas

• Monthly quotas increase from 2,000 to 3,000 calls on Starter and from 10,000 to 20,000 on Pro, without a price increase.
• Legacy 1,000- and 1,500-call plans move to 3,000; 15,000-call plans move to 20,000.
• Existing customer prices remain locked in, and their quotas will not be reduced. See plans and billing for current allowances.

​

Logo Link

• Logo Link is available: display a company’s logo using a URL built from its domain.

​

Updated guides

• Updated setup instructions and task guides make it easier to find the right API and make a first request.

​

Context.dev for Make

• Connect Context.dev to Make to build automations with other apps.

​

More accurate social links

• Brand retrieval no longer returns generic social-network homepages as company profiles.

​

Dashboard redesign

• The developer dashboard has a new layout and navigation for API tools and account management.

​

More playground tools

• Email-based brand lookup and font extraction are now available in the dashboard playground.

​

Faster brand retrieval

• Brand retrieval is faster for domains that have not already been cached.

​

Prefetch by work email

• Prefetch brand data from a work email before you need the result.
• The operation costs 0 credits for paid subscribers and rejects free or disposable email addresses with 422.

​

Fewer duplicate backdrops

• Backdrop processing removes duplicate images from brand results.

​

Merchant phone numbers

• Include a merchant phone number in transaction enrichment to help verify the company match.

​

Request brand updates

• Request a brand-data update for incorrect logos, colors, or company details.
• The launch offers review within 24 hours.

​

More context for transaction matching

• Improve merchant matching with an optional category code (mcc), city, and country (country_gl).
• See Transaction enrichment.

​

Billing on successful requests

• At this release, billing changed to charge only successful 200 responses, leaving invalid inputs, inaccessible domains, and rejected email types unbilled.
• See current billing guidance for today’s policy.

​

Clearer demo errors

• Demos explain invalid inputs more clearly, and website links and missing-page handling have been corrected.

​

Sign in with GitHub

• Sign up or sign in to Context.dev with GitHub.

​

Brand lookup by email

• Look up a brand from a work email.
• The API extracts the company domain and returns a specific error for free or disposable email providers.

​

More brand lookup options

• Brand lookup documentation now covers domain, company-name, and ticker inputs with a consistent response shape.
• See Get brand data.

​

Faster API responses

• API requests return faster across endpoints.

​

Fewer false logo matches

• Logo detection rejects more decorative icons that could otherwise be mistaken for company logos.

​

International ticker exchanges

• Set ticker_exchange to distinguish the same ticker across international exchanges.
• Without it, the lookup assumes a US listing.
• See Look up by ticker.

​

Zapier 2.0

• Zapier 2.0 adds three actions, updates brand retrieval, and includes new recipes.
• See the Zapier guide for the current action catalog.

​

API History beta

• The dashboard’s new API History beta shows request activity and consumption trends.

​

Corrected usage charts

• Fixed incorrect displays in the dashboard’s request-history charts.

​

Faster uncached brand lookups

• Brand retrieval returns results faster for uncached domains.

​

Styleguide controls

• Styleguide adds a prioritize choice between speed and quality.
• Styleguide removes this parameter in April 2026; follow the current design-system guide.

​

Faster styleguide extraction

• Styleguide extraction returns results faster.

​

Company website links

• Brand profiles add a links object for discovered careers, blog, terms, privacy, contact, and pricing pages.

​

Starter pricing

• Starter pricing changes for new customers.
• Existing customers retain their previous plan rates.

​

Download invoices

• View and download invoices from the dashboard.

​

One site for docs and account tools

• Account management and documentation now share the main website, removing the need to switch to a separate developer-portal domain.

​

Change plans directly

• Upgrade or downgrade a plan without cancelling it first.

​

Richer brand previews

• Brand previews now include contact email, phone, industries, and backdrops.

​

Industry and subindustry labels

• Brand profiles add Easy Industry Classification in industries.eic, with industry and subindustry labels.

​

Look up by name or ticker

• Look up brand profiles by company name or stock ticker, in addition to domain.
• See Get brand data for supported lookup methods.

​

More wordmark coverage

• Improved wordmark extraction and refreshed brand records increase the coverage of wide brand logos.

​

Keep history when rotating keys

• Rotating an API key no longer clears request history.
• Usage remains attached to the organization, including historical calls.

​

Adult-content classification

• Brand profiles add is_nsfw to help applications identify adult-content websites.

​

Higher-resolution logos

• Logo extraction prioritizes higher-resolution sources and filters low-quality variants.

​

Original search API retired

• The original search endpoint is retired and returns 410 Gone.
• This historical retirement does not apply to the Web Search and Brand Search APIs introduced in 2026.

​

Company names and dashboard feedback

• Updates on July 13 and 19 improve extracted company names and add a dashboard feedback widget for reporting data issues and suggestions.

​

Clearer API errors

• Invalid endpoints and request parameters now return clearer errors, with improved traffic protection for service availability.

​

Typography through Styleguide

• Use Styleguide for detailed typography, including families, weights, sizes, line heights, and spacing.
• It provides more typography detail than the Brand response.

​

Faster screenshots and styleguides

• Screenshot and Styleguide requests are faster for both cached and newly captured pages.

​

Screenshot API beta

• The beta Screenshot API captures viewport or full-page images and returns a hosted image URL.

​

Styleguide API beta

• The beta Styleguide API extracts colors, typography, spacing, shadows, and component styles for application theming.

​

Developer portal and affiliates

• The developer portal has updated navigation and documentation.
• The original affiliate program launches with 20% commission; a new program replaces it in June 2026.
• See current affiliate terms.

​

Faster brand retrieval

• On-demand brand retrieval returns results faster.

​

Brand prefetching

• Paid subscribers can prefetch brand data before it is needed, at 0 credits, to reduce the wait for a later lookup.

​

Speed and timeout controls

• Use maxSpeed for faster Brand lookups with less detail, and timeoutMS to set a request deadline.
• Timed-out requests return 408. See Timeouts for the current replacement syntax and limits.

​

Request-history filters

• Filter dashboard request history by recent periods, longer date ranges, or all available history.

​

First Zapier integration

• The first Zapier integration launches with a brand-retrieval action.

​

Ruby SDK and image metadata

• The official Ruby SDK is available.
• Image metadata also adds has_opaque_background for assets with their own background.

​

Python SDK

• The official Python SDK is available.

​

Typed TypeScript SDK

• The rewritten TypeScript SDK exposes request and response types for easier integration.

​

Faster brand lookups

• Brand lookups return results faster.