Firecrawl changelog: what's new each month
Every stable Firecrawl release summarised by month: the highlights, new features, improvements, fixes and anything you need to act on. 12 months covered; the current month updates daily.
June 2026
1 release: 2.11.0
June brought major expansions to Firecrawl's research capabilities, access options, and data quality features. The month focused on making the platform more accessible, intelligent, and useful for AI-driven workflows.
Highlights
- Launched Firecrawl Research Index with 3M+ arXiv papers and their GitHub code for AI/ML research discovery.
- Keyless access now available for core endpoints through official MCP, CLI, and SDK clients.
- Automatic PII redaction strips sensitive data like names, emails, and phone numbers from scraped content.
- Monitors gained field-level JSON diffs and LLM-powered goal filtering to reduce false alerts.
- Video discovery now works on any page, not just known providers, with full metadata extraction.
New
- Firecrawl Research Index: search 3M+ arXiv papers and their associated GitHub code, issues, and PRs with daily refreshes.
- Keyless access for /scrape, /search, /interact, and /parse endpoints via official MCP, CLI, and SDK clients.
- Automatic PII redaction option to strip personal and sensitive data from scraped content.
- deterministicJson format that generates structured JSON without running an LLM per request, with per-site schema caching.
- Video discovery on any page type, returning URL, title, thumbnail, duration, and more.
- CDP WebSocket URL access to control live Firecrawl browser sessions with Playwright, Puppeteer, or other automation tools.
- LLM-based goal evaluation for monitors to filter meaningful changes from noise.
- Field-level JSON diffs for monitor comparisons to show exact field value changes between runs.
Improved
- Video format expanded to find and extract video metadata from any page, not limited to known providers.
- Monitor alerts now use LLM judgment to filter changes against user-defined goals, reducing irrelevant notifications.
- Browser session management enhanced with direct access to Chrome DevTools Protocol for advanced automation.
- JSON monitoring improved with field-level comparison instead of rendered page diffs for precision tracking.
May 2026
1 release: 2.10
May brought significant expansion to Firecrawl's capabilities, adding local file parsing, advanced query modes for scraping, and new search refinements. The release focuses on making extracted data more useful and controllable for AI applications.
Highlights
- Upload and parse local files (PDF, DOCX, XLSX, and more) up to 50 MB directly to the /parse endpoint.
- Use natural-language questions with /scrape to get grounded, hallucination-free answers from web pages.
- Lockdown Mode isolates scraping to Firecrawl's index with zero outbound requests and zero data retention.
- Extract video URLs from supported sites like YouTube with authenticated download support.
- Filter /search results by domain with new includeDomains and excludeDomains parameters.
New
- /parse endpoint supports uploading local files in PDF, DOCX, DOC, ODT, RTF, XLSX, XLS, and HTML formats up to 50 MB.
- question format returns grounded, hallucination-free answers from scraped content using a managed model chain with fallback.
- highlights format returns exact sentences, code blocks, and table rows matching your query, with consecutive items rejoined.
- video format scrapes signed downloadable video URLs from supported sites with cookie forwarding for authenticated access.
- includeDomains and excludeDomains parameters on /search to scope results to specific sites.
- /search feedback endpoint to rate results and earn credit refunds.
- robotsUserAgent parameter to evaluate robots.txt rules against a custom user agent string.
- Lockdown Mode on /scrape for serving results exclusively from Firecrawl's index with zero outbound requests.
Improved
- Table and reading order preservation improved in /parse output.
- Custom robots.txt evaluation now supports separate customRobotsAgent org flag for crawl requests.
April 2026
1 release: 2.9.0
April brought significant expansion to Firecrawl's scraping and interaction capabilities, including browser automation, multiple output formats, and new SDK support. The release adds more control over content extraction and introduces specialized parsers for different content types.
Highlights
- Browser interaction now available via the /interact endpoint to click buttons, fill forms, and extract dynamic content from pages.
- Query format lets you pass natural-language prompts to get direct answers extracted from scraped pages.
- PDF parsing modes (fast, auto, ocr) and page limits give you control over extraction depth and OCR behavior.
- New Java and Elixir SDKs provide official support for v2 API across more development environments.
- onlyCleanContent parameter automatically strips navigation, ads, and cookie banners from markdown output.
New
- Browser interaction via /interact endpoint for clicking, form-filling, and navigation on live pages
- Query format parameter on /scrape endpoint to get direct natural-language answers from page content
- Audio format option for scrape responses with audio output as a document field
- onlyCleanContent parameter to remove navigation, ads, and non-content elements from markdown
- PDF parsing modes selection (fast, auto, ocr) with maxPages option for extraction control
- Official Java SDK with full v2 API support
- Official Elixir SDK with full v2 API support
- Dedicated Wikimedia engine for improved Wikipedia and Wikimedia page scraping
Improved
- Legacy .doc file parsing now supported alongside modern formats
- PDF extraction offers configurable depth control through mode and page limit settings
- Wikimedia content extraction optimized for improved output quality
February 2026
1 release: 2.8.0
February brought significant expansion to Firecrawl's agent capabilities, introducing parallel processing, a command-line interface, and three new specialized models to handle different types of extraction and research tasks.
Highlights
- Parallel Agents now enables running thousands of agent queries simultaneously using the new Spark 1 Fast model.
- New Firecrawl CLI provides command-line access to scrape, search, crawl, and map functionality.
- Firecrawl Skill allows AI agents like Claude Code and Codex to autonomously use Firecrawl within their workflows.
- Three new models for the /agent endpoint: Spark 1 Fast for instant retrieval, Spark 1 Mini for research queries, and Spark 1 Pro for advanced extraction.
- Agent features now include webhooks, model selection, and new MCP Server tools.
New
- Parallel Agents capability for simultaneous large-scale /agent queries
- Firecrawl CLI tool for all core functions
- Firecrawl Skill for AI agent integration
- Spark 1 Fast model for instant retrieval
- Spark 1 Mini model for complex research queries
- Spark 1 Pro model for advanced extraction tasks
- Webhooks for agent workflows
- MCP Server tools for agents
Improved
- Agent enhancements including model selection capabilities
- Search execution performance optimized
- Redis call optimization for faster platform performance
- SDK compatibility with Zod v4
December 2025
1 release: 2.7.0
December brought enterprise search capabilities, improved content extraction accuracy, and a new experimental scraping system alongside various performance enhancements.
Highlights
- ZDR Search support is now available for enterprise customers.
- Screenshots are now faster and more accurate.
- Partner Integrations API entered closed beta testing.
- Branding detection and extraction were improved significantly.
- Redirect handling was enhanced for more reliable crawling.
New
- ZDR Search support for enterprise users
- Partner Integrations API (closed beta)
- NOQ Scrape System (experimental)
Improved
- Branding Format detection
- Branding Extract capability
- Screenshot speed and accuracy
- Redirect handling
- Self-hosted deployment
November 2025
1 release: 2.6.0
November brought a unified billing system merging credits and tokens, along with full support for Firecrawl's branding format across all platforms. The release focused on performance improvements and reliability across endpoints.
Highlights
- Credits and tokens unified into a single billing system, with extract now using credits at a 15:1 token ratio.
- Branding format now fully supported in Playground, MCP, and both JavaScript and Python SDKs.
- All endpoints significantly faster with improved reliability and error handling.
- Change tracking now detects web page content updates faster and more reliably.
- Instant credit purchases available directly from the dashboard.
New
- Instant credit pack purchases from dashboard without waiting for auto-recharge.
- Unified billing model combining credits and tokens into single system.
Improved
- Markdown parsing and conversion accuracy enhanced.
- Main content extraction accuracy improved.
- All endpoints performance increased with improved reliability.
- Change tracking detection speed and reliability improved.
- Error handling improved across the system.
Fixed
- Change tracking issues resolved.
- PDF timeout issues fixed.
- Bug fix on v2 experimental.
October 2025
2 releases: 2.4.0 → 2.5.0
October brought significant infrastructure improvements and new search capabilities to Firecrawl. The month saw the introduction of a next-generation search API, enhanced observability features, and expanded file format support.
Highlights
- New x402 search endpoint offers improved accuracy and speed for searching.
- PDF-specific search category lets you filter search results to PDFs only.
- Excel file scraping now supported alongside existing file formats.
- Crawl status reporting is now more accurate and real-time.
- RabbitMQ event system enables event-based job communication and better job prefetching.
New
- PDF search category for v2/search endpoints
- Gemini 2.5 Flash CLI image editor integration
- x402 next-generation search API endpoint
- RabbitMQ event-based communication system for jobs
- Excel (.xlsx) file scraping support
- Per-owner and per-group concurrency limiting
- Team flag to skip country checks
- Test-site app for improved CI testing
Improved
- Crawl Status API now provides more accurate and real-time reporting
- OpenTelemetry distributed tracing significantly enhanced for better service observability
- Request-level metrics added for both Scrape and Search endpoints
- Group backlog handling and group operations improved
- Crawl architecture redesigned with NUQ concurrency tracking
- Dynamic concurrency calculation implemented
- NUQ metrics now always populated for improved observability
Fixed
- Low-results and robots.txt warnings now clearly reported to users when crawls are limited
September 2025
2 releases: 2.2.0 → 2.3.0
September brought significant performance improvements and expanded capabilities, including MCP version 3 support, YouTube transcript extraction, and major speed gains for document parsing. The month focused on stability, new integrations, and better support for enterprise and self-hosted deployments.
Highlights
- MCP version 3 is now stable with HTTP Transport and SSE modes.
- Map feature is now 15x faster and supports more URLs.
- YouTube transcripts can now be extracted directly.
- Docx parsing is 50x faster.
- Webhooks now support signatures, extract data, and event failure tracking.
New
- YouTube transcript extraction support.
- ODT and RTF file parsing.
- Webhook signatures and event failure reporting.
- maxPages parameter for PDF parsing in the v2 scrape API.
- Queue status endpoint.
- Enterprise auto-recharge support.
- Support for additional scraping locations: Canada, Czech Republic, Israel, India, Italy, Poland, and Portugal.
- NuQ worker deployment example for Kubernetes.
Improved
- Map performance increased 15x with broader URL support.
- Search reliability improved.
- Docx parsing speed increased 50x.
- Usage tracking is now organized by API Key.
- Self-hosted deployment experience improved significantly.
- MCP now compatible with versions 1, 2, and 3.
Fixed
- Timeout race condition resolved.
- Infinite scrape loop fixed.
- Location query bug resolved.
August 2025
3 releases: 2.0.0 → 2.1.0
August brought a major version 2 release with faster caching defaults and a new summary format, followed by security and capability improvements. The month focused on expanding search capabilities and enhancing content extraction.
- v1 client must be replaced with v2 client; initialize with new Firecrawl({ apiKey }) syntax.
- JSON extraction format changed: use { type: "json", prompt, schema } instead of old "extract" format; old format renamed to "json".
Highlights
- Requests now cache by default with a 2-day maximum age, and sensible defaults like ad blocking and image compression are enabled.
- New summary format lets you get concise page summaries directly without additional processing.
- Search now covers news, images, and specific domains like GitHub and academic research sites beyond standard web results.
- JSON extraction uses a cleaner object format with optional schema support and natural-language prompt guidance.
- Crawling now accepts natural-language prompts to automatically determine paths and limits.
New
- Summary format option to receive direct page content summaries.
- Search across news and images sources in addition to web results.
- Search categories parameter to filter results by topic.
- GitHub source for searching repositories, code, issues, and documentation.
- Research source for academic and research websites including arXiv, Nature, IEEE, and PubMed.
- Image extraction support on the v2 scrape endpoint.
- Data attribute scraping to extract data- HTML attributes.
- Google Drive scraping for TXT, PDF, and Sheets files.
Improved
- Screenshot options now use object form for better control over full-page, quality, and viewport settings.
- Crawling now handles hash-based routes.
- Google Drive scraping expanded to support multiple file types.
Fixed
- SSRF vulnerability via malicious webhook in self-hosted version.
July 2025
2 releases: 1.14.0 → 1.15.0
July brought enterprise features and reliability improvements to Firecrawl, including authenticated scraping, SSO support, and enhanced scraping performance. The month saw over 50 bug fixes and improvements across the platform.
Highlights
- Authenticated scraping capability now available (waitlist).
- SSO for enterprise customers added.
- Zero data retention option for enterprise users.
- Improved scraping reliability and p75 speeds.
- OMCE support added to scrapeURL and HTML transformer.
New
- Authenticated scraping for accessing protected content.
- Zero data retention option for enterprise deployments.
- SSO authentication for enterprise accounts.
- Search params now included in activity logs.
- createdAt field added to /crawl/active response.
- OMCE support in scrapeURL and HTML transformer.
- New MCP version with maxAge parameter and improved tool calling.
Improved
- P75 scraping speeds improved.
- Overall scraping reliability enhanced.
- Proxy usage optimization.
- Logging improved to include search parameters and cache age.
- MCP tool calling behavior improved.
Fixed
- JSON format issue on search results.
- Over 50 bug fixes and improvements merged.
June 2025
4 releases: 1.10.0 → 1.13.0
June brought major enhancements to Firecrawl's web scraping and search capabilities, including a new web search API, the Firecrawl Index for faster scrapes, and expanded format support. The month focused on giving users more control over crawling behavior and output options.
Highlights
- Search the web and retrieve full content from results in a single API call with customizable parameters and output formats.
- Firecrawl Index launched to speed up scrapes up to 5x for opted-in users.
- New concurrency system lets you control maximum concurrency per request for better resource management.
- Expanded scraping support now includes Google Docs, Google Slides, and PDF generation from pages.
- Improved crawl control with domain-wide crawling and subdomain crawling capabilities.
New
- Web search API that combines search results with full content extraction in one call.
- Firecrawl Index for faster scraping performance when opted in.
- Per-request concurrency control system.
- Google Docs scraping support.
- Google Slides scraping support.
- PDF generation from current page.
- Subdomain crawling with allowSubdomains parameter.
- Stealth Mode support for Australia, France, and Germany.
Improved
- Timeout handling and content type improvements for scrapeURL and PDF operations.
- Redis support with separate non-eviction Redis configuration.
- Activity Logs with better FIRE-1 request support and webhook event viewing.
- Crawl parameter now allows following sibling and parent URLs, not just child paths.
- Search functionality with ignoreBlockedURLs and concurrency limit options.
- Auto mode proxy for scraping across scrapeURL and JavaScript SDK.
- /search Playground enhanced with location parameters.
- Screenshot quality configurable with higher resolution options available.
Fixed
- Vertex AI provider bug fixed with updated model references.
- Improved handling of various content types in scraping operations.
- Better concurrency limit management for search requests.
May 2025
1 release: 1.9.0
May brought significant performance improvements and better LLM provider support to Firecrawl. The release focused on speed, reliability, and easier configuration across the platform.
Highlights
- Crawling is now much faster with global adoption of a cacheable lookup system.
- Fixed support for LLM providers to work correctly with extraction and processing.
- Added crawl delay support with per-crawl concurrency limiting for controlled resource usage.
- SDK 2.0 now includes change tracking to monitor modifications.
- Improved setup process and documentation for parameters and prompts.
New
- Crawl delay support with per-crawl concurrency limiting
- Change tracking in SDK 2.0
Improved
- Crawl performance significantly faster
- LLM provider integration and compatibility
- Setup and configuration process
- Prompt guidance and examples
- Parameter usage documentation
- Supabase client handling
Fixed
- Supabase client issues
- LLM provider support
Summaries are written automatically from the official release notes (full changelog ↗); check the original notes before relying on a detail. Firecrawl: pricing, features and alternatives · All changelogs