A multi-agent Claude Code plugin that transforms raw CSV, Excel, and JSON files into insights, reports, dashboards, and business recommendations — fully automated.
Drop your data anywhere. Point the plugin at it. Get a complete analysis — cleaned data, statistical insights, charts, an interactive dashboard, and a business-ready report.
Give any command a project name or a folder path from anywhere on your system. The plugin auto-copies data files into the project registry.
5 specialist agents (Data Engineer, Statistician, Visualizer, Reporter, Strategist) coordinate through an orchestrated pipeline with parallel execution.
Profiling, cleaning, EDA, chart generation, dashboard building, and report writing — all happen automatically with domain-specific intelligence.
Open your interactive dashboard, read the Markdown report, explore the charts. Optionally send results to Slack, Gmail, or Discord via MCP.
From profiling to production monitoring, the plugin covers the entire data analysis lifecycle.
5 specialist agents work in parallel. The Statistician and Dashboard builder run simultaneously — cutting total analysis time nearly in half.
Auto-discover Shopify, databases, Google Sheets, Slack, Gmail, Discord, and more. Pull live data in, push results out — all automated.
Standalone HTML dashboards with Chart.js. KPI cards with delta indicators, responsive layout, works offline after first load. Open in any browser.
Automatically detects E-Commerce, Marketing, Healthcare, or General data from column names and runs domain-specific analysis patterns.
Watch your data with CronCreate scheduling. Re-profile every N minutes. Get alerts via Slack or webhook when quality drops.
All 18 commands work with Claude Code voice mode. Say "analyze my sales project" and the pipeline runs. Natural speech routing built-in.
Analyze 10+ projects in parallel with worktree isolation. Each project gets its own agent. Combined cross-project insights at the end.
3-attempt auto-recovery: try, auto-fix (install deps, fix paths, handle encoding), WebSearch for unknown errors. Rarely needs manual intervention.
Every instruction is explicit step-by-step. Haiku runs as reliably as Opus. Use fast models for simple tasks, powerful models for deep analysis.
Every command accepts a project name or a full filesystem path. Results always go to output/<project>/.
| Command | Description | Model |
|---|---|---|
:analyze | Full 5-agent swarm pipeline — profile, clean, EDA, visualize, report, dashboard | Sonnet |
:profile | Data profiling and quality assessment | Haiku |
:clean | Data cleaning and transformation (swarm mode for 10+ files) | Haiku |
:query | Ask natural language questions about your data | Sonnet |
:visualize | Generate charts and visualizations | Haiku |
:report | Comprehensive Markdown analysis report | Sonnet |
:dashboard | Standalone interactive HTML dashboard | Sonnet |
| Command | Description | Swarm Pattern |
|---|---|---|
:watch | Live-monitor with CronCreate — re-profile, re-dashboard, or full re-analysis on schedule | Loop Swarm |
:batch-analyze | Parallel agents per project in isolated worktrees (up to 10 concurrent) | Worktree Swarm |
:compare | Side-by-side dataset comparison with parallel profiling | Worktree Pair |
:research | Multi-source web research: Reddit, X, HN, YouTube, web (6 parallel searches) | Search Swarm |
| Command | Description | Model |
|---|---|---|
:debug | Auto-diagnose pipeline failures, check deps, match error patterns, search fixes | Sonnet |
:schedule | Schedule future/recurring tasks via CronCreate | Haiku |
:notify | Configure webhook notifications for pipeline events | Haiku |
:simplify | 3-agent code review swarm (Reuse + Quality + Efficiency) | Sonnet |
:api | Export all artifacts as structured JSON, optionally serve on port 8080 | Haiku |
| Command | Description | Model |
|---|---|---|
:connect | Auto-discover & configure MCP data sources and messaging apps | Haiku |
:live-update | Send results to connected apps (Slack, Gmail, Discord, Telegram) | Haiku |
See how teams and founders use 10x-Analyst-Loop to turn raw data into decisions — without writing a single line of code.
| File | Contains |
|---|---|
| data-profile.md | 50,234 orders, 97.1% quality |
| insights.json | 12 structured findings |
| dashboard.html | KPI cards, trends, product breakdown |
| report.md | Executive summary + recommendations |
| charts/ | 8 PNG: revenue, products, AOV, RFM, cohort |
campaign, clicks, impressions, ctr, conversion| # | Project | Rows | Quality |
|---|---|---|---|
| 1 | hr-data | 12,450 | 94.2% |
| 2 | finance-q1 | 8,320 | 98.1% |
| 3 | support-tickets | 45,000 | 87.3% |
| 4 | product-usage | 120,400 | 92.7% |
| 5 | inventory | 3,200 | 99.4% |
MCP (Model Context Protocol) servers extend the plugin with external data sources, messaging, and actions. Auto-discovered at runtime — zero hardcoding.
:connect runs 9+ parallel ToolSearch calls to find all available MCPs
Discovered tools are classified and saved to .mcp-config.json
Pipeline stages auto-pull data and push results through connected MCPs
An orchestrated pipeline of specialist agents with parallel execution at key stages for maximum throughput and token efficiency.
order revenue price product customercampaign clicks impressions ctr conversionpatient diagnosis treatment medicationTrack tokens, costs, active agents, scheduled jobs, and context usage — all in your terminal. 5 themes, 4 layouts.
One command. Five agents. Zero config. Works on Haiku, Sonnet, and Opus.