v2.1.1 — Agent Swarms + MCP Integration

Data Analysis on
Autopilot

A multi-agent Claude Code plugin that transforms raw CSV, Excel, and JSON files into insights, reports, dashboards, and business recommendations — fully automated.

# Point at any folder. Walk away. Come back to a full analysis.
/10x-analyst-loop:analyze C:/Users/you/sales-data

# Connect Shopify + Slack for live data + notifications
/10x-analyst-loop:connect my-store shopify
/10x-analyst-loop:connect my-store slack #analytics

# Analyze all projects in parallel — one agent per project
/10x-analyst-loop:batch-analyze input/*
18
Slash Commands
5
Specialist Agents
8
Swarm Patterns
12+
MCP Integrations

Four Steps. Zero Config.

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.

1

Point at Data

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.

2

Agent Swarm Activates

5 specialist agents (Data Engineer, Statistician, Visualizer, Reporter, Strategist) coordinate through an orchestrated pipeline with parallel execution.

3

Analysis Runs

Profiling, cleaning, EDA, chart generation, dashboard building, and report writing — all happen automatically with domain-specific intelligence.

4

Results Delivered

Open your interactive dashboard, read the Markdown report, explore the charts. Optionally send results to Slack, Gmail, or Discord via MCP.

Everything You Need for Data Analysis

From profiling to production monitoring, the plugin covers the entire data analysis lifecycle.

Multi-Agent Swarm

5 specialist agents work in parallel. The Statistician and Dashboard builder run simultaneously — cutting total analysis time nearly in half.

🔌

MCP Integration

Auto-discover Shopify, databases, Google Sheets, Slack, Gmail, Discord, and more. Pull live data in, push results out — all automated.

📊

Interactive Dashboards

Standalone HTML dashboards with Chart.js. KPI cards with delta indicators, responsive layout, works offline after first load. Open in any browser.

🔬

Domain Auto-Detection

Automatically detects E-Commerce, Marketing, Healthcare, or General data from column names and runs domain-specific analysis patterns.

🕑

Live Monitoring

Watch your data with CronCreate scheduling. Re-profile every N minutes. Get alerts via Slack or webhook when quality drops.

🎤

Voice-Friendly

All 18 commands work with Claude Code voice mode. Say "analyze my sales project" and the pipeline runs. Natural speech routing built-in.

🚀

Batch Processing

Analyze 10+ projects in parallel with worktree isolation. Each project gets its own agent. Combined cross-project insights at the end.

🛠

Self-Healing Pipeline

3-attempt auto-recovery: try, auto-fix (install deps, fix paths, handle encoding), WebSearch for unknown errors. Rarely needs manual intervention.

🤖

Model-Agnostic

Every instruction is explicit step-by-step. Haiku runs as reliably as Opus. Use fast models for simple tasks, powerful models for deep analysis.

18 Commands. One Plugin.

Every command accepts a project name or a full filesystem path. Results always go to output/<project>/.

Core Analysis 7 commands
CommandDescriptionModel
:analyzeFull 5-agent swarm pipeline — profile, clean, EDA, visualize, report, dashboardSonnet
:profileData profiling and quality assessmentHaiku
:cleanData cleaning and transformation (swarm mode for 10+ files)Haiku
:queryAsk natural language questions about your dataSonnet
:visualizeGenerate charts and visualizationsHaiku
:reportComprehensive Markdown analysis reportSonnet
:dashboardStandalone interactive HTML dashboardSonnet
Power Commands 4 commands
CommandDescriptionSwarm Pattern
:watchLive-monitor with CronCreate — re-profile, re-dashboard, or full re-analysis on scheduleLoop Swarm
:batch-analyzeParallel agents per project in isolated worktrees (up to 10 concurrent)Worktree Swarm
:compareSide-by-side dataset comparison with parallel profilingWorktree Pair
:researchMulti-source web research: Reddit, X, HN, YouTube, web (6 parallel searches)Search Swarm
System & DevOps 5 commands
CommandDescriptionModel
:debugAuto-diagnose pipeline failures, check deps, match error patterns, search fixesSonnet
:scheduleSchedule future/recurring tasks via CronCreateHaiku
:notifyConfigure webhook notifications for pipeline eventsHaiku
:simplify3-agent code review swarm (Reuse + Quality + Efficiency)Sonnet
:apiExport all artifacts as structured JSON, optionally serve on port 8080Haiku
Integration 2 commands
CommandDescriptionModel
:connectAuto-discover & configure MCP data sources and messaging appsHaiku
:live-updateSend results to connected apps (Slack, Gmail, Discord, Telegram)Haiku

Real Problems. Real Solutions.

See how teams and founders use 10x-Analyst-Loop to turn raw data into decisions — without writing a single line of code.

Use Case 1
E-Commerce Revenue Intelligence
"Where is my revenue going?" — Shopify Store Owner
You run a Shopify store with 50,000+ orders. You export data monthly but never have time to analyze it. You need revenue trends, product performance, customer segmentation, and cohort retention — all automated.
# Connect Shopify for live data pulls /10x-analyst-loop:connect my-store shopify # Run full analysis /10x-analyst-loop:analyze my-store # Follow up /10x-analyst-loop:query my-store "highest CLV segments?" /10x-analyst-loop:watch my-store 1h --full /10x-analyst-loop:live-update my-store slack
  • Revenue trend analysis (+23% MoM) with growth drivers identified
  • Top 10 products by revenue — 3 products drive 60% of total
  • RFM segmentation: Champions 12%, At Risk 8%
  • Cohort retention heatmap: Month 3 drop-off at 45%
  • Interactive dashboard + Markdown report + 8 PNG charts
Shopify MCP Slack MCP

Output Artifacts

FileContains
data-profile.md50,234 orders, 97.1% quality
insights.json12 structured findings
dashboard.htmlKPI cards, trends, product breakdown
report.mdExecutive summary + recommendations
charts/8 PNG: revenue, products, AOV, RFM, cohort
Use Case 2
Cross-Platform Campaign Performance
"Which campaigns actually work?" — Marketing Team Lead
Your team runs campaigns across Google Ads, Meta, and email. You have CSVs from each platform but no unified view. You need cross-platform ROAS, funnel analysis, and budget allocation insights.
# Drop all platform exports into one folder /10x-analyst-loop:analyze C:/Marketing/Q1-campaigns # Compare quarters /10x-analyst-loop:compare q1-campaigns q2-campaigns # Custom chart /10x-analyst-loop:visualize q1-campaigns "ROAS by channel" # Schedule weekly report every Monday 9am /10x-analyst-loop:schedule q1-campaigns report "every monday 9am"
  • Domain auto-detected as MARKETING from column names
  • Unified cross-platform view: Google Ads 3.2x ROAS vs Meta 1.8x
  • Email has highest conversion (4.2%) but lowest reach
  • Funnel analysis with drop-off rates per channel
  • Budget reallocation recommendations with expected impact

Domain Detection

Trigger columns found:
campaign, clicks, impressions, ctr, conversion

Domain: MARKETING

Auto-runs: Campaign performance, Channel ROAS, Funnel analysis, Budget efficiency
Compare output:
comparison-vs-q2-campaigns.md
Row counts, quality diffs, metric trends, quarter-over-quarter insights
Use Case 3
Bulk Dataset Analysis at Scale
"I need to analyze 15 datasets by Friday" — Data Analyst
You received 15 project datasets from different departments. Each needs profiling, cleaning, and a basic report. Doing them one by one would take days.
# Analyze ALL projects in parallel — one agent each /10x-analyst-loop:batch-analyze input/* # Or specify individual projects /10x-analyst-loop:batch-analyze hr-data finance-q1 support-tickets
  • 15 parallel agents in isolated worktrees — one per project
  • Each runs the full :analyze pipeline independently
  • Combined summary table with cross-project insights
  • Total time = roughly one analysis (parallel execution)
  • Failed projects can be retried independently

Batch Results

#ProjectRowsQuality
1hr-data12,45094.2%
2finance-q18,32098.1%
3support-tickets45,00087.3%
4product-usage120,40092.7%
5inventory3,20099.4%
+ 10 more projects...
Use Case 4
Automated Data Quality Monitoring
"Alert me when data quality drops" — Ops Manager
You have an automated pipeline that dumps CSVs into a folder. Sometimes the export breaks and produces garbage data. You need to know the moment it happens — not days later when dashboards break.
# Set up Slack alerts /10x-analyst-loop:connect pipeline-data slack #data-alerts # Watch every 10 minutes /10x-analyst-loop:watch pipeline-data 10m --profile # Also notify PagerDuty via webhook /10x-analyst-loop:notify pipeline-data https://events.pagerduty.com/v2/enqueue
  • Re-profiles data every 10 minutes via CronCreate
  • Compares quality score against baseline
  • Slack alert when quality drops >5%: files affected, issue details
  • Webhook POST to PagerDuty with structured JSON payload
  • Full change log in watch-log.md for auditing
Slack MCP Webhook CronCreate

Alert Example (Slack)

⚠️ Data Quality Alert — pipeline-data

Quality dropped from 96.4% to 82.1%
3 new issues detected:
• orders.csv: 14% missing in customer_id
• orders.csv: 230 duplicate rows
• products.csv: encoding error

— 10x Analyst Loop
Use Case 5
Market Research + Internal Data Analysis
"Research the market and analyze our data" — Startup Founder
You're building a SaaS product. You need to research competitor pricing trends AND analyze your own user engagement data — then share findings with your product team via Slack and investors via email.
# Part 1: Market research (no data files needed) /10x-analyst-loop:research "SaaS pricing trends 2026" # Part 2: Analyze your own data /10x-analyst-loop:analyze user-engagement # Part 3: Distribute findings /10x-analyst-loop:connect user-engagement slack #product /10x-analyst-loop:connect user-engagement gmail /10x-analyst-loop:live-update user-engagement slack /10x-analyst-loop:live-update user-engagement gmail
  • 6 parallel WebSearches: Reddit, HN, X, YouTube, general web
  • Cross-platform convergence detection (recurring themes)
  • Full engagement analysis: cohort retention, feature usage, churn
  • Auto-send insights to Slack #product and investors via Gmail
  • Research brief saved to output/research/saas-pricing-trends-2026.md
WebSearch Slack MCP Gmail MCP

Research Output

saas-pricing-trends-2026.md

Key Findings:
1. Usage-based pricing up 34% YoY (Reddit + HN consensus)
2. PLG companies shifting to hybrid models
3. AI-native SaaS charging per outcome, not per seat
4. Free tier becoming table stakes for dev tools
Live Update Sent:
Slack #product + Gmail investors@company.com
✓ Delivered

Connect Everything. Automate Everything.

MCP (Model Context Protocol) servers extend the plugin with external data sources, messaging, and actions. Auto-discovered at runtime — zero hardcoding.

Data Sources
  • Shopify
  • PostgreSQL / Supabase
  • Google Sheets
  • Notion
  • Airtable
  • Stripe
Messaging
  • Slack
  • Gmail
  • Discord
  • Telegram
  • Microsoft Teams
Actions
  • Composio
  • Zapier
  • Webhooks
🔍

Discover

:connect runs 9+ parallel ToolSearch calls to find all available MCPs

📌

Configure

Discovered tools are classified and saved to .mcp-config.json

Automate

Pipeline stages auto-pull data and push results through connected MCPs

5-Agent Swarm Pipeline

An orchestrated pipeline of specialist agents with parallel execution at key stages for maximum throughput and token efficiency.

User Request (text or voice) | v +-------------------+ | ORCHESTRATOR | <-- INDEX.md routes everything | (Skill Router) | +--------+----------+ | +------+---------------+---------------+----------+ v v v v v +---------+ +---------+ +----------+ +--------+ +----------+ | Data | | Stats | |Visualizer| |Reporter| |Strategist| | Engineer| | ician | | | | | | | +----+----+ +----+----+ +----+-----+ +---+----+ +----+-----+ | | | | | v v v v v Clean & EDA & Charts & Markdown Business Profile Stats Dashboard Report Actions

Auto-Detected Domains

🛒

E-Commerce

Columns: order revenue price product customer
  • Revenue trends & MoM growth
  • Top products by revenue
  • Average Order Value (AOV)
  • RFM customer segmentation
  • Cohort retention analysis
📈

Marketing

Columns: campaign clicks impressions ctr conversion
  • Campaign performance comparison
  • Channel ROAS analysis
  • Funnel analysis & drop-offs
  • Budget allocation efficiency
🏥

Healthcare

Columns: patient diagnosis treatment medication
  • Outcome distributions
  • Treatment comparisons
  • Time-to-event analysis
📊

General Tabular

Any other column patterns
  • Correlation matrix (flag |r| > 0.7)
  • Distributions & group-by aggregations
  • Anomaly detection (IQR method)
  • Pareto analysis (80/20 rule)

Real-Time Session Monitoring

Track tokens, costs, active agents, scheduled jobs, and context usage — all in your terminal. 5 themes, 4 layouts.

Skill: Agent │ GitHub: user/repo/main Model: Claude Opus │ Dir: projects/my-data Tokens: 45k in + 12k out │ Cost: $0.32 ($0.04/m) Session: 57k total (8 calls) $0.0400/call +120/-15 2m30s Cache: W:12k R:33k @main Agents: ●●● (3) │ Cron: (1) ▰▰▰▰▰▯▯▯ HIGH 15% remaining Context: ████████████████████████████████░░░░░░░░ 85%
default nord tokyo-night catppuccin gruvbox

Turn Raw Data Into Decisions

One command. Five agents. Zero config. Works on Haiku, Sonnet, and Opus.

/10x-analyst-loop:analyze your-data


View on GitHub Visit 10x.in