Changelog
What shipped, and when.
Every release, newest first. Features, improvements and fixes are marked so you can skim for the kind you care about.
v1.15.0
October 22, 2026
New- FixedActivating a tab now makes it the tab your next actions land on. Before, activate moved only the foreground tab and the viewer, so switching back to the first of two tabs and then clicking clicked in the second
- FixedListing tabs reports the tab untargeted actions really land on as activeTargetId — it used to name the oldest tab while actions went to the newest
- BetterThe active-tab choice is shared across the API's machines, so it holds on every request rather than only the one that made it
- BetterThe API reference now documents listing and creating tabs beside closing and activating them
v1.14.0
October 2, 2026
New- NewCAPTCHAs clear themselves: a challenge on a page a session just opened is solved automatically, and the navigation reports what it cost
- NewCapSolver joins 2Captcha and Anti-Captcha as a solving provider, with automatic fallback between whichever are configured
- NewSpend caps on automatic solving, per session and per hour, counted across machines — and solving stops entirely if the counter cannot be reached
- BetterScraping jobs solve CAPTCHAs too, which the per-source setting never actually switched on
v1.13.0
September 30, 2026
New- NewWebhooks for saved logins: get told the moment a site signs a profile out, and when a browser session ends, instead of finding out on the next run
- NewA session can opt out of automatic profile saving with autoCapture: false, for runs that should leave a saved login exactly as it was
- FixedAn automatic save can no longer replace a working login with a signed-out page
- FixedSaved page storage is kept per origin, so two sites on one host no longer share it, and older profiles keep what they already held
v1.12.0
September 24, 2026
Major- NewClaude can drive a session directly: post the calls from Claude's browser toolset to one endpoint and send back what it returns, with tabs, screenshots and element refs handled for you
- NewA human-approval gate for Claude's actions — clicks, typing, navigation or file uploads can stop and wait for a person to approve or deny before they run
- NewClaude can attach a file to a page's upload field
- NewSelect a dropdown option by its visible label, press real key chords such as Ctrl+A, and get an event for every file a page downloads
- BetterScreenshots can be capped to a longest edge, so they fit a model's image limits at any count
- FixedEach session's downloads are its own, a script that returns a promise now returns what it resolves to, and navigation settles on the page it is going to rather than the one it left
v1.11.0
September 21, 2026
Major- NewTake over a session from your phone: one finger is the mouse, two fingers pan and zoom, and a keyboard button types into the remote page
- NewA tab frozen by a native dialog now says so: the error names the dialog, you can accept or dismiss it (including one that opens behind another), and a stuck tab can be replaced without losing the sign-in
- NewChallenge pages are noticed as they appear, and a refused action says which field and which rule refused it
- BetterStarting a cloud browser from cold takes about a quarter of the time it did
- BetterThe live viewer always shows the whole page, sized to the panel it is in, and narrow panels receive sharper frames
- BetterMinutes in which a browser could not respond are no longer billed, and a browser that stops responding restarts itself and resumes its sessions
- FixedEnter submits a form from the viewer, and the first click after a viewer connects is no longer lost
v1.10.0
August 26, 2026
Major- NewRead a page the way a screen reader does — roles, names, and state — then click or type against a ref instead of guessing a CSS selector that breaks on the next re-render
- NewRun up to 25 browser actions in one call, search a page for a control by describing it, and fill an entire form in a single request
- NewRecord a browser session to video — an MP4 of everything the session did, downloadable while the session is still alive
- NewNavigation can wait for the page's own scripts to finish running, and a new wait-for-function blocks until any condition you name comes true
- NewCoordinate and path drags, wheel scrolling into nested panes, modifier clicks, and paste — including pasting an image into an editor that has no upload button
- NewA browser session carrying a login can hold a static exit IP, and a dead one is detected and cleared automatically
- BetterThe live viewer captures at Retina density and streams binary frames, so the remote page looks sharp and responds faster
- BetterThe AI chat agent's browser tools now match the API's, including press-and-hold for human-verification widgets
- FixedClosed the last cases where one browser session could see another's tabs, pages, or cookies
- FixedA credit purchase shows up immediately instead of waiting out a cache, and file uploads no longer report success for a file the page never received
v1.9.0
August 22, 2026
New- NewUpload a file from your own machine into a live browser session — the remote file dialog is replaced with your browser's own picker
- NewEmbedded integrations can scope every call to one of their own end users, so one customer can never reach another's session
- BetterA browser session now survives its underlying machine being replaced, rebuilding itself with the same cookies, tabs, and exit IP
- BetterCredit balances load faster — the balance read no longer sits in front of every deduction
- FixedClicks in the live viewer land where you aimed instead of roughly 9% too low, and the page is no longer squeezed into 800x600
- FixedYou are charged only for browser time we can prove was live, and waived charges leave a visible trace
- FixedCredit purchase webhooks reach Scrappy again — deliveries had been landing on a path that did not exist
- FixedLive progress streams reconnect after a laptop sleeps instead of quietly giving up
v1.8.0
August 15, 2026
New- NewCloud browser sessions can detect a CAPTCHA and, with a solver key configured, clear it
- NewWeb search is now its own API and MCP tool — structured results billed in credits, with no search provider key of your own
- NewAn API key can be bound to a single project, and the binding is visible, enforced, and editable
- BetterThe project ID is now visible and copyable where you need it for API calls
- FixedEach browser session's timezone is derived from its own proxy exit IP, so it no longer contradicts where the traffic appears to come from
- FixedRepaired 60+ MCP tool parameters that were silently dropped before reaching the API
- FixedOutreach no longer reports success for mail that was never sent
v1.7.0
July 24, 2026
Major- NewCloud browser sessions are metered per minute, capped by per-key quotas, and embeddable on your own domain with short-lived viewer tokens
- NewBrowser profiles — sign in once and every future session starts already authenticated, cookies and local storage intact
- NewThe browser API reached parity with the in-app agent: drag and drop, wait for the network to settle, block requests by type or URL, and pierce shadow DOM
- NewChoose how a page is reduced before extraction — full, simplified, or text-only — so an AI reads structure instead of markup
- NewBrowser sessions are scoped to the active project, and the session panel shows the proxy and live egress IP
- BetterA browser session's lifetime is based on idle time with a keepalive, replacing the hard age cap that killed sessions mid-use
- BetterCloud browser sessions present a coherent fingerprint — real Chrome, matching platform and version, no spoofed WebGL
- BetterThe interactive viewer is responsive, and copy/paste works inside it
- FixedClosed AI billing gaps, including unbounded free-tier chat usage
v1.6.0
July 20, 2026
New- NewHosted MCP endpoint at /api/mcp — external AI agents authenticate with a Scrappy API key, drive the cloud browser, and can converse with the Scrappy agent itself
- NewThe AI assistant can now search and read Scrappy's own documentation when answering you
- FixedLong chat sessions no longer overflow the model's context limit
- FixedThe cloud browser wakes reliably from suspend and reports an outage as an outage instead of a generic error
v1.5.0
June 29, 2026
New- BetterSecurity hardening across access control, SSRF, prompt injection, and encryption — every finding from a deep audit closed
- BetterMoved to self-hosted Redis and browser machines that suspend when idle, cutting running costs without changing behavior
- BetterA consistent visual pass across the dashboard — shared loading, alert, and status components, and colors that work in both light and dark mode
v1.4.0
May 28, 2026
New- NewPut a browser session to sleep and wake it later with its tabs and cookies intact, instead of losing it
- BetterBrowser screenshots render as images in chat instead of raw data, and oversized tool results are trimmed
- FixedZombie, stale, and orphaned browser sessions are detected and recovered rather than lingering
v1.3.0
April 11, 2026
Major- NewCloud browser — run several isolated browser sessions at once and watch or drive any of them live from the dashboard
- NewConsole, network, and DOM panels for a live browser session, streamed from the browser itself
- NewCloud browser SDK — drive a Scrappy browser from your own code with an API key, including an embeddable live-view React component
- NewAutonomous pipeline — set a lead target and Scrappy discovers sources, scrapes, enriches, and validates until it hits the number
- NewImport leads and datasets straight from Airtable, Google Sheets, and folk.app with smart column mapping
- NewEmail finder — find a person's email from a name and company, verified before it is returned
- NewGenerate and download real spreadsheets from your data, with formula-injection protection
- BetterLinkedIn enrichment runs through the anti-detection browser with full session cookies, and can force a re-enrich
- BetterEmail validation adds NeverBounce as a default tier, per-lead SMTP control, and SOCKS5 proxy support
v1.2.0
March 31, 2026
Major- NewMCP server — drive Scrappy from Claude Code and any MCP client, with the full toolset exposed over stdio
- NewThe AI chat agent can now reach every part of Scrappy — 73 new tools across chat, projects, outreach, LinkedIn, GDPR, enterprise, and admin
- NewDatasets — collect any structured data from the web with AI extraction, then export as JSONL or a HuggingFace dataset
- NewContact form outreach — Scrappy detects a contact form on a lead's site, fills it, submits, and tracks the result
- NewWorkflow automation, team collaboration, and advanced analytics — triggers, schedules, approvals, mentions, delegation, and funnel/cohort/forecast reporting
- NewPause and resume a running scraping job, and retry a stuck one without paying for it twice
- NewMove or copy leads between projects, and transfer or duplicate leads and sources between users
- NewBuild a campaign from a brief — upload a deck, add instructions, and keep your progress if you close the tab
- NewLink a GitHub account to an existing password login and sign in either way
- BetterPer-site stealth scraping with proxy selection, fingerprint evasion, and WAF bypass tuned per source
- BetterScraping survives a flaky browser — circuit breaker, per-URL retry, health checks, and automatic restart
- BetterFewer junk leads — role-address detection, a minimum quality threshold, and an AI cleanup pass you can run in bulk
- BetterShift-click to select a range of leads, and your page, sort, and column preferences survive a reload
v1.1.0
March 24, 2026
New- NewAPI key authentication on all endpoints — use Bearer tokens for programmatic access
- BetterRefactored large service files into focused single-responsibility modules for better maintainability
- FixedPersist Playwright browsers in Docker volume to avoid redownloading on container restart
- FixedResolved GitHub OAuth sign-in issues caused by missing User table columns
- FixedFixed DB migration drift, middleware rename, and worker rebuild performance
v1.0.0
March 4, 2026
Major- NewTeam features — organizations with roles, API key management, source templates marketplace (SSO and SCIM are planned, not available)
- NewProxy profiles — residential and datacenter proxy configuration for scraping
- NewEmail suppression lists — compliance-focused suppression management
- NewData destinations — sync leads to external systems via webhooks, databases, and CRMs
- NewSource lead targets — a weekly lead target per source, with an alert when it is missed
- NewSelector healing — auto-repair broken CSS selectors when page structure changes
- NewAI chat agent — Claude-powered assistant for managing leads, sources, and jobs via natural language
- NewSource health analytics dashboard with alert escalation rules
- NewPipeline analytics and admin metrics with LTV tracking
- Better270+ API endpoints with comprehensive documentation
- BetterConcurrency and correctness audit — rate limiter hardening, pagination overflow protection, timing-safe comparisons
v0.9.0
February 26, 2026
New- NewDeal campaign builder — discover sources, rank accounts, and find profile commonalities using AI
- NewAccounts dashboard — aggregated view of companies scraped across all leads
- NewLinkedIn commonalities analyzer — surface shared traits across profiles using Perplexity AI
- NewCommercial construction lender profile type template
- BetterCampaign wizard with guided deal parameter configuration
v0.8.0
February 17, 2026
New- NewCredits system with balance tracking and checkout via cred.diy
- NewWebhooks with HMAC signing, delivery history, and one-click retry
- NewAngelSend integration — export leads, send bulk emails, and generate AI-powered sequences
- NewEmail warmup domains — 5-week progressive ramp with DNS tracking and auto-pause on failures
- BetterPer-user AngelSend API keys with fallback to global environment variable
v0.7.0
February 17, 2026
New- NewGDPR tools — consent records, data export, and account deletion
- NewLinkedIn credential manager with health tracking, rate-limit detection, and soft-ban detection
- NewAccount lockout after failed login attempts with lockout status endpoint
- NewData retention worker for expired leads, exports, and audit logs (enforcement is being rolled out)
- BetterAudit logging with a 2-year retention period and sensitive field sanitization
- BetterSession max-age reduced to 1 day for improved security
v0.6.0
February 17, 2026
New- NewProfile Types system — define custom entity schemas with JSON Schema field definitions
- NewMatching Engine — filter and score profiles with FILTER, SCORED, and AI rule types
- NewAI matching powered by Perplexity sonar-pro with customizable prompt templates
- NewProfile export with optional rule-based filtering, match scores, and AI reasoning columns
- BetterMulti-tier email validation (DNS, MX, SMTP) with batch processing
v0.5.0
February 17, 2026
New- NewLead CSV import with column mapping UI and duplicate detection
- NewBulk lead actions — delete, tag, and export in CSV or JSON
- NewManual lead creation via Add Lead dialog
- NewScraping job detail view with per-URL progress, retry count, and cancellation
- BetterAutomatic lead deduplication using canonical SHA-256 hashing
- FixedSoft-delete for leads — deleted leads are hidden but recoverable
v0.4.0
February 17, 2026
New- NewProjects — organize sources and leads into separate workspaces
- NewSource health monitoring — HEALTHY / WARNING / CRITICAL badges with dismissable alerts
- NewSelector validator — test CSS selectors against live pages before activating a source
- NewAlternative source suggestions for CRITICAL sources
- BetterSources now start as DRAFT; explicit activation required
v0.3.0
February 1, 2026
New- NewAI-powered source discovery using Perplexity sonar-pro
- NewDeep crawl engine with configurable URL frontier and pagination detection
- NewSitemap discovery and URL pattern matching
- BetterRedis-backed rate limiting with sliding window per endpoint
- BetterStandardized API error handling and route middleware
v0.2.0
January 15, 2026
New- NewBackground worker with BullMQ for scraping, email validation, and retention jobs
- NewCredential manager for storing login credentials used during scraping
- NewProxy rotation support via Webshare API
- BetterCloud browser integration for reliable, bot-resistant scraping
v0.1.0
January 1, 2026
Launch- NewInitial launch — web scraping sources, lead management, and email validation
- NewNextAuth authentication with project-scoped access
- NewPostgreSQL + Prisma ORM with Redis queue
- NewDashboard with lead list, source management, and job monitoring
Start
See it for yourself. 100 credits, no card.
Enough for ten scrapes or a first hour in the cloud browser. No card, no call, no minimum.
The open webOne session away