What's New in 100xBot - December 2025: The Month We Made Browser Automation Feel Human
TL;DR: December brought the biggest update to 100xBot yet—171 commits, 67,000+ lines of new code, and a complete rethinking of how you interact with your AI browser assistant. We introduced granular permissions, a full PostgreSQL database running in your browser, native PDF and archive processing, and a smarter agent that remembers and learns.
The Headline: Your Browser Just Became a Supercomputer
This month, we asked ourselves: What would it take for 100xBot to handle enterprise-level tasks without ever leaving your browser?
The answer was ambitious. We needed to give users complete control over what the agent can do. We needed professional-grade document processing. We needed SQL databases. We needed the agent to be smarter about when to act and when to ask.
And we built all of it.
- Trust, But Verify: The New Permission System
The Problem We Solved:
Users told us they loved 100xBot's power but sometimes felt nervous about giving an AI unrestricted access to their browser. "What if it clicks something I didn't want?" "Can I trust it on my banking site?"
What Changed:
100xBot now asks before it acts. Every capability—clicking, typing, navigating, extracting data—can be individually approved or denied.
How It Works:
When 100xBot needs to perform an action (like clicking a button or extracting text), you'll see a permission dialog that lets you:
- Allow or Deny the specific action
- Remember your choice for 1 hour, 1 day, 1 week, or forever
- Scope your permission to this exact page, this path, or the entire domain
- Bulk manage permissions when multiple requests queue up
Why This Matters for Users:
You're now in complete control. Allow 100xBot full access on your internal tools, but require approval on sensitive sites. Trust builds over time—the agent learns your preferences and stops asking for things you've already approved.
Why This Matters for Enterprises:
Compliance teams can audit exactly what actions 100xBot has permission to perform. IT admins can set baseline policies. This is browser automation with governance built in.
- PostgreSQL in Your Browser: The PGLite Revolution
The Problem We Solved:
Users were extracting data from websites, but then asking "Now what?" They wanted to analyze, filter, join, and transform data—but had to export to spreadsheets or external databases.
What Changed:
100xBot now includes a full PostgreSQL database running entirely in your browser using WebAssembly. No server. No setup. Just SQL.
What You Can Do Now:
- Store scraped data in proper relational tables
- Run complex SQL queries with JOINs, aggregations, window functions
- Fuzzy text search using pg_trgm extension ("find products similar to 'widjet'")
- Persistent storage across sessions via IndexedDB
- Extensions included: pg_trgm (fuzzy search), fuzzystrmatch (phonetic matching), uuid-ossp, hstore (key-value), ltree (hierarchical data), tablefunc
Real Example:
"Scrape the top 100 products from Amazon, store them in a table, then show me the 10 most expensive items in the Electronics category with more than 4 stars"
This now happens entirely in your browser. The agent extracts the data, stores it in SQL, and runs the analysis query—all without sending your data anywhere.
Why This Matters:
Data stays local. Analysis happens instantly. No API limits. No external dependencies. You can build dashboards, track changes over time, and run analytics on any website data—powered by real SQL.
- Professional Document Processing: PDF and Archives
The Problem We Solved:
Users often work with downloaded files—PDFs, ZIP archives, compressed data. Previously, 100xBot could download files but couldn't do much with them.
What Changed:
We integrated MuPDF (the PDF engine behind Firefox) and 7-Zip as WebAssembly modules, giving 100xBot native document processing capabilities.
PDF Tools (MuPDF)
- Extract text from any PDF (entire document or specific pages)
- Render pages to images at any DPI
- Get metadata (author, creation date, page count)
- Search text within PDFs
- Merge multiple PDFs into one
- Split PDFs into individual pages
- Handle encrypted PDFs with password support
Real Example:
"Download this invoice PDF, extract the total amount and due date, then store it in my invoices table"
Archive Tools (7-Zip)
- Read: 7z, ZIP, RAR, TAR, GZ, BZ2, XZ, CAB, ISO, and 20+ more formats
- Create: 7z, ZIP, TAR, GZ archives
- Compression levels from store-only to ultra compression
- Password protection with encrypted filenames (7z format)
- Selective extraction using patterns
Real Example:
"Download this data export, extract only the CSV files, then import them into my database"
Why This Matters:
Document processing traditionally required external services, desktop apps, or complex server setups. Now it happens in your browser tab in seconds. Privacy-conscious users can process sensitive documents without uploading them anywhere.
- Smarter Workflow Recording: Watch, Learn, Repeat
The Problem We Solved:
Recording browser actions to create automations was powerful but mysterious. Users weren't sure what was being captured or how to get the best results.
What Changed:
We completely redesigned the recording experience with:
Recording Explainer Dialog
Before recording starts, a clear 3-step guide explains:
- Perform your actions normally
- Click "Stop Recording" when done
- 100xBot analyzes and creates reusable workflows
The tip: "Keep actions focused on a single task for better workflow quality"
Context-Aware Recording
The interaction recorder now captures context around elements, not just the clicks. When you interact with a button, we record:
- The button's semantic role and labels
- Surrounding text that identifies it
- Multiple fallback selectors for resilience
- The page state and URL patterns
Result: Workflows that work reliably even when websites change their layouts slightly.
Sidepanel-Independent Recording
Recording now continues even if you close the sidepanel. Work naturally across tabs without worrying about losing your recording session.
- 100 Ready-to-Use Examples: Inspiration at Your Fingertips
The Problem We Solved:
New users often asked: "What can 100xBot actually do?" They needed inspiration, not documentation.
What Changed:
The new Empty State Guide presents 100 example automations organized by category:
- Video & Media Generation (10 examples): Create video ads, animated GIFs, time-lapses, AI voiceovers
- Data Visualization & Dashboards (10 examples): Convert tables to charts, build KPI dashboards, create heatmaps
- Diagram & Flowchart Generation (10 examples): Generate Mermaid diagrams, mind maps, decision trees
- Form & Data Automation (10 examples): Record form fills, bulk submissions, test data generation
- PDF & Document Generation (10 examples): Compile catalogs, generate invoices, create ebook PDFs
- Image Processing (10 examples): Batch background removal, watermarks, format conversions
- Data Extraction & Analysis (10 examples): Extract emails, phone numbers, prices, sentiment analysis
- Content Transformation (10 examples): Translation, summarization, social media repurposing
- Automation & Workflows (10 examples): Pagination handling, infinite scroll, multi-tab operations
- Advanced Features (10 examples): AI chatbots, accessibility audits, SEO analysis
Each example shows capability tags (what tools it uses) and one-click execution to try immediately.
- A Redesigned Interface: Faster, Cleaner, More Responsive
The Problem We Solved:
As we added features, the UI needed to scale. Users wanted faster interactions and cleaner visuals.
What Changed:
Shadcn/UI Migration
We moved from Chakra UI to Shadcn/ui, a modern component system that gives us:
- Smaller bundle size (faster loading)
- Better dark mode support
- More accessible components
- Consistent design language
Sidebar Redesign
The primary and secondary sidebars were completely rebuilt:
- Collapsible panels for more workspace
- Quick actions for common operations
- Better organization of workflows, data, history, and permissions
Content Truncation
Large outputs (like scraping 1000 items) no longer freeze the UI. Content is intelligently truncated with expand options.
Delete Confirmations
Destructive actions now have proper confirmation dialogs, preventing accidental data loss.
- Background Task Management: Async Everything
The Problem We Solved:
Long-running operations (large file processing, complex workflows) would block the agent from responding. Users couldn't start new conversations while waiting.
What Changed:
Any capability can now run in the background with __async: true. The agent:
- Starts the operation immediately
- Returns control to the conversation
- Updates results when complete
- Supports cancellation if you change your mind
New Capabilities:
- List running tasks: See all background operations and their status
- Get task status: Check progress of specific operations
- Cancel tasks: Stop operations that are no longer needed
- Task counts: Dashboard-style overview (running, completed, failed)
Why This Matters:
Start a large PDF extraction, then ask 100xBot something else while it processes. The agent manages multiple operations concurrently without blocking your workflow.
- Haiku Default: Faster, Cheaper, Smarter
The Problem We Solved:
Users were paying for more AI compute than many tasks required. Quick questions didn't need the most powerful model.
What Changed:
Claude Haiku is now the default model for:
- Quick responses
- Simple extractions
- Routine operations
You can still switch to Sonnet or Opus for complex tasks. The result: faster responses and lower costs for everyday use.
Per-Conversation Model Selection:
Each conversation can use a different model. Start with Haiku for exploration, switch to Opus for complex analysis—all in the same session.
- Token Management: Conversations That Don't Break
The Problem We Solved:
Long conversations would eventually hit context limits, causing confusion or errors. Users had to start new sessions to continue.
What Changed:
Auto-Compaction
When conversations approach the token limit, 100xBot automatically compacts history while preserving:
- Key decisions and results
- Important memories
- Current task context
Smart Truncation
Tool results that would overflow the context are now truncated intelligently:
- Large tables show summaries
- File contents show relevant excerpts
- Counts and metadata always included
Token Visibility
You can now see token usage in real-time, understanding when compaction will trigger.
- Enterprise-Grade Testing: Quality You Can Trust
Behind the Scenes:
We added 5,400+ new automated tests this month:
| Component | Test Count |
|---|---|
| Grep Engine | 1,044 tests |
| File Storage | 958 tests |
| PGLite Database | 722 tests |
| Archive Processing | 409 tests |
| Message Filtering | 384 tests |
| SSE Parsing | 186 tests |
| Array References | 125 tests |
Plus new integration tests for Document Service and File Sync.
Why This Matters:
Every feature is verified across edge cases. When we say PDF extraction works, it's been tested against hundreds of document types. When we say SQL queries are safe, we've verified injection prevention.
The Numbers
| Metric | December 2025 |
|---|---|
| Commits | 171 |
| Files Changed | 387 |
| Lines Added | 67,604 |
| Lines Removed | 35,172 |
| New Capabilities | 6 |
| New Test Cases | 5,400+ |
| Contributors | 2 |
What's Next
December laid the foundation. In January, we're focused on:
- Scheduled Automations: Run workflows on timers
- Multi-Tab Orchestration: Coordinate actions across browser windows
- Visual Debugging: Step through workflows with screenshots
- Community Workflows: Share and import automations
Try It Now
100xBot is available at https://100x.bot. The December update is live for all users.
Questions? Feature requests? Visit our https://github.com/100x-bot/100xbot to join the conversation.
100xBot: Your browser, supercharged.