Guerilla Marketing Team

cd project/promotion/ && npx xswarm-buzz

Ditch the Marketing Team. You ARE the Marketing Team.

One developer + AI agents = unstoppable marketing force

Stop wasting 20 hours a week on social media. Talk to AI for 20 minutes, let it handle the grinding. You stay in flow, building stuff that matters.

Part of the xSwarm family: Maximum capability. Zero overhead. Team of One.

The Problem

You're a developer. You built something awesome. Now you need to market it.

Your options:

1. Hire a marketing team → $$$$$, meetings about meetings, "let's circle back on this synergy"
2. Do it yourself manually → 4 hours/day on social media, soul-crushing repetition, spreadsheet hell
3. Use xswarm-buzz → AI agents do the grinding, you stay in flow, 20 min/week

You know which one makes sense.

The xSwarm Philosophy: Be the Team You Want to Lead

Think "Army of One" but for developers. Maximum capability, minimum headcount.

The xSwarm Family (xswarm.ai):

• 🎯 xSwarm (npx xswarm) → Your AI planning & coding team
• 🎙️ xSwarm-Boss (Rust, voice-first) → Your AI personal assistant managing everything
• 📣 xswarm-buzz (npx xswarm-buzz) → Your AI marketing department ← YOU ARE HERE

One developer = One complete company. That's the vision.

How It Actually Works (Pure Conversation)

No dashboards. No forms. No "configure your settings". Just talk to AI like you'd talk to your marketing manager.

$ npx xswarm-buzz

> "What should I do today?"

📊 MORNING BRIEFING:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
• 3 hot leads replied overnight (all positive 🔥)
• Mississippi Coaches campaign: 87% acceptance rate
• Reddit post got 234 upvotes, 12 signups
• Today's mission: 15 LinkedIn outreach, 1 post, 8 comment replies

> "Let's do it. Generate today's outreach."

🎯 Analyzing 47 prospects...
🧠 Reading your context (product, voice, successful patterns)...
✨ Generating 15 personalized messages...
📤 Ready to review (no templates, all fresh AI generation)

> "Show me message 1"

[AI shows generated message for review]

> "Looks good. Approve all 15."

✅ Messages approved and queued.
⏱️  Time spent: 4 minutes (including your review)
📊 You manually: would've taken 4 hours

You stay in control. AI does the heavy lifting.

That's the whole interface. Conversational. Natural. Fast.

Real Scenarios: What This Looks Like

Scenario 1: Launch a New Campaign

> "I built an app for high school coaches to manage compliance.
   Need to reach 100 of them in Mississippi."

🔬 Researching market...
   • Found 247 coaches on LinkedIn in MS
   • 14 active Facebook groups for HS coaches
   • Pain points: state compliance, paperwork overload

📝 Building persona...
   • Age 35-55, overwhelmed with admin work
   • Posts about "drowning in paperwork"
   • Responds well to specific compliance mentions

🎯 Strategy generated:
   • Platform mix: LinkedIn (primary), Facebook groups (secondary)
   • Hook: "Cut compliance paperwork by 80%"
   • 30-day cadence: 5 outreach/day, 3 posts/week, engagement

📅 Campaign "Mississippi Coaches Q4" ready to launch.
   Launch now or review strategy first?

> "Launch it."

🚀 Campaign active.
   Day 1 outreach queued.
   I'll handle the daily execution.
   Check in anytime: "How's Mississippi Coaches doing?"

Scenario 2: Daily Marketing Grind (Automated)

> "Do today's marketing"

⚡ Running daily workflow...

✅ LinkedIn: 15 connection requests sent
   (each personalized using prospect's recent posts + our context)

✅ Posted to Twitter, LinkedIn, Product Hunt
   (variations tailored to each platform's culture)

✅ Responded to 12 comments across platforms
   (maintaining your voice, helpful not salesy)

✅ Updated 3 active campaigns with today's metrics

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⏱️  Done in 4 minutes.
📊 Manual equivalent: 4 hours of your life
💰 Saved: $200 of your time (at $50/hr)

Back to building.

Scenario 3: Learning & Optimization

> "Why is Campaign B crushing Campaign A?"

🧠 Analyzing performance data...

📊 Campaign A: Generic Value Props
   • Acceptance rate: 22%
   • Message angle: "Save time with our tool"
   • Response: meh

📊 Campaign B: Specific Pain Points
   • Acceptance rate: 43%
   • Message angle: "Eliminate state compliance headaches"
   • Response: "OMG yes where do I sign"

🔍 Pattern detected:
   Mentioning "state compliance" specifically = 2.1x response rate
   Generic "save time" = everyone says this, ignored

💡 Applying learning to all active campaigns...
   ✅ Updated 3 campaigns with compliance focus
   ✅ Regenerating queued messages
   ✅ Adjusting content calendar

Your marketing just got smarter while you were coding.

Why This Beats Manual Marketing

Time Effectiveness

Translation: You save 19 hours and 40 minutes per week to actually build stuff.

Coordination

You manually:

• Spreadsheets tracking which platform you posted to
• Calendar reminders for outreach cadence
• "Wait, did I already message this person?"
• Copy-pasting between platforms
• Forgetting to follow up

xswarm-buzz:

• AI tracks everything with perfect memory
• Coordinates timing across all platforms
• Never duplicates outreach
• Auto-adapts content per platform
• Follows up automatically

Result: Your sanity remains intact.

Personalization: Templates vs. AI Generation

Template approach (what everyone does):

Hi [FIRST_NAME],

I noticed [COMPANY] on LinkedIn and thought you might be interested in...

Best regards,
[YOUR_NAME]

Everyone knows. Nobody responds. 8% acceptance rate.

xswarm-buzz approach:

> Generate message for prospect ID 47

🧠 Analyzing prospect...
   • Name: Mike Johnson, Athletic Director at Madison HS
   • Recent post: "Drowning in compliance paperwork again"
   • Engagement pattern: Responds to specific solutions
   • Language: Direct, no-BS style

✨ Generated message:

"Mike - saw your post about compliance paperwork. Been there.

Built something that auto-fills the MHSAA forms using your
roster data. Coaches in Tupelo cut their compliance time
from 6 hours to 30 minutes weekly.

Worth a quick call?"

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
This message:
• References his actual pain point from his own words
• Mentions his state's specific organization (MHSAA)
• Uses nearby city (Tupelo) for credibility
• Matches his direct communication style
• Specific time savings (not generic "save time")

Result: 43% acceptance rate. 2-3x better than templates.

The difference? AI reads context, analyzes the prospect, generates fresh. Every. Single. Time.

Human-in-the-Loop: You're Always in Control

Critical: AI suggests. You approve. Always.

xswarm-buzz is NOT an autonomous spam bot. It's your AI assistant that drafts messages for you.

The workflow:

1. AI generates personalized message suggestions
2. You review each message (or batch review)
3. You can edit, approve, or reject
4. Nothing sends without your explicit approval
5. AI learns from your edits and choices

You maintain full control. The AI just makes the drafting 10x faster.

Think of it like: AI is your copywriting intern. You're still the marketing director who reviews and approves everything before it goes out.

Learning System: Gets Better Over Time

You manually:

• "I think this is working?"
• Gut feelings
• Maybe some Excel tracking
• No pattern recognition
• Starting from scratch each time

xswarm-buzz:

• Tracks every interaction automatically
• Identifies winning patterns ("compliance" keyword = 2.1x response)
• A/B tests messaging approaches
• Learns your successful voice
• Auto-applies learnings to future campaigns

Result: Week 1 = 25% acceptance rate. Week 8 = 45% acceptance rate. Same effort, better results.

Context-Driven = Actually Smart Marketing

Most marketing tools are dumb. They don't know your product, your market, your voice. They're glorified schedulers.

xswarm-buzz reads context from your project folder before doing anything.

Your Project/promotion/
├── context/
│   ├── product.md          ← What you built, why it matters
│   ├── market.md           ← Who needs it, market landscape
│   ├── personas.md         ← How your users think and talk
│   └── voice.md            ← How YOU talk (AI mimics this)
│
├── campaigns/
│   ├── active/             ← Running campaigns
│   ├── planned/            ← Queued up
│   └── completed/          ← Historical learnings
│
├── audiences/              ← Prospects, contacts (gitignored)
├── interactions/           ← Engagement history (gitignored)
└── .claude/skills/         ← Marketing capabilities

Every action starts with: "Let me read your context first..."

Result:

• Messages sound like you
• Targeting actually makes sense
• Strategy aligns with your product
• Learning compounds over time

This is what "AI-native" actually means.

The Tech (For Developer-Marketers)

You're a developer. You care how this works.

Multi-Model Architecture

Not locked into one LLM. Use whatever makes sense:

• Claude Sonnet (strategy, creative writing, reasoning)
• Claude Haiku (data extraction, simple analysis)
• GPT-4 / GPT-3.5 (if you prefer OpenAI)
• Gemini Pro / Flash (Google's models)
• Local models (Llama 3.2, Phi, etc. for privacy/cost)

xswarm-buzz picks the cheapest sufficient model for each task.

Cost Optimization: Subagent Orchestration

Three-tier system:

Example workflow:

1. Campaign strategy → Tier 1 (Claude Sonnet) - needs reasoning
2. Prospect list extraction → Tier 3 (Local Llama) - simple parsing
3. Message generation → Tier 1 (Claude Sonnet) - needs creativity
4. Scheduling → Tier 3 (Local model) - trivial logic
5. Performance analysis → Tier 2 (Claude Haiku) - moderate analysis

Result: You get Tier 1 quality where it matters, pay Tier 3 prices where it doesn't.

You Control the Model Preferences

Config file: config/model-preferences.json

{
  "tier1": {
    "preferred": ["claude-sonnet", "gpt-4"],
    "maxCostPerRequest": 0.50
  },
  "tier2": {
    "preferred": ["claude-haiku", "llama-3.2-local"],
    "maxCostPerRequest": 0.10
  },
  "tier3": {
    "preferred": ["llama-local", "phi-local"],
    "preferLocal": true
  },
  "embeddings": {
    "provider": "local",
    "model": "all-MiniLM-L6-v2"
  }
}

System respects your preferences. Want everything local for privacy? Set preferLocal: true. Have OpenAI credits? Put GPT-4 first. Optimize for cost? Bias toward local models.

Skills-Based Architecture

Marketing capabilities are modular Skills (folders containing instructions + scripts + resources).

Locations:

• ~/.claude/skills/ → Your personal global skills
• .claude/skills/ → Project-specific skills (shared via git)
• npm bundled → Core skills included

Core Skills (bundled with npm package):

• campaign-creator → Interactive campaign builder with market research
• message-generator → AI-generated personalized messages
• prospect-analyzer → Deep prospect intelligence from public data
• daily-briefing → Morning routine with prioritized actions
• skill-fetcher → Download community skills from GitHub

Integration Skills (fetch on-demand):

• linkedin-connector → OAuth, search, profile analysis, messaging
• gmail-connector → Email sending, thread management
• facebook-connector → Group discovery, post scheduling
• twitter-connector → Posting, thread creation, engagement
• reddit-connector → Subreddit targeting, comment strategies

Analytics Skills:

• metrics-tracker → Campaign performance, conversion rates
• learning-capture → Analyze outcomes, improve generation
• ab-test-analyzer → Statistical analysis of message variants

Workflow Skills:

• outreach-executor → Daily outreach with message generation
• content-scheduler → Manage content calendar across platforms
• response-handler → Generate replies to engagement

Community Skills (100+ on GitHub):

• Tweet generators, email sequencers, meme creators, sentiment analyzers, competitive monitors, etc.
• Fetch with: > "Install the tweet-generator skill"

Create your own: Use built-in skill-creator to build custom capabilities. Share on GitHub.

Privacy & Security

Local data storage:

• Prospects, interactions, auth tokens → All stored locally
• .gitignore pre-configured for sensitive files
• Optional encryption for backups

Optional local models:

• Run everything on your machine (Llama, Phi, etc.)
• Zero data leaves your computer
• Free (after hardware cost)
• Good for Tier 2/3 tasks

Auth management:

• OAuth tokens in config/auth/ (gitignored)
• System keychain integration (macOS Keychain, Windows Credential Manager)
• Per-skill authentication (LinkedIn, Gmail, etc.)

Machine-Controllable: JSONL Protocol

Two modes:

Human mode (default): Natural conversation

npx xswarm-buzz
> "What should I do today?"

Machine mode (for automation):

npx xswarm-buzz --json

JSONL protocol for external orchestration (e.g., xSwarm-Boss controlling this).

Message types: command, approval, status, progress, result, log, error, control

Example:

{"type": "command", "action": "daily_briefing"}
{"type": "status", "message": "Analyzing overnight activity..."}
{"type": "result", "data": {"leads": 3, "acceptance_rate": 0.87}}

This lets other tools (like xSwarm-Boss) control xswarm-buzz programmatically.

Installation & Quick Start

Install

npm install -g xswarm-buzz

Or use without installing:

npx xswarm-buzz

First Time Setup (10-15 minutes)

cd your-project
mkdir promotion && cd promotion
npx xswarm-buzz init

Two-phase AI-powered setup:

Phase 1: AI Backend (Required)

• Option 1: Use Claude Code (recommended - no API key needed if logged in)
• Option 2: Enter Anthropic API key (get free credits at console.anthropic.com)
• Option 3: Use OpenAI API key
• Option 4: Use local Ollama (free, runs on your machine)

Phase 2: Conversational Interview

• AI asks smart questions about your product
• Analyzes your website if provided
• Builds personalized marketing strategy
• Creates campaign files ready to execute
• Configures promotional channels

What gets created:

• context/ → Product, market, personas, voice profiles
• campaigns/ → Ready-to-launch campaigns tailored to your goals
• channels/ → Recommended marketing channels pre-configured
• knowledge/ → Quick reference guide
• config/.env → AI backend + API keys configured
• .claude/skills/ → skill-creator for building custom automation

Start Marketing

npx xswarm-buzz

That's it. Start talking:

> "I need to reach 50 potential users this month. My product is [describe]."

> "Create a LinkedIn campaign targeting [persona]."

> "What should I do today?"

> "Generate 10 personalized outreach messages."

> "How are my campaigns performing?"

Conversational interface = No manual to read.

Success Metrics: Team-of-One Marketing Department

Target performance (what a $120k/year marketing manager would deliver):

Daily Metrics

• ✅ 15 outreach messages sent (>30% acceptance rate)
• ✅ 1 post published across platforms
• ✅ 20%+ reply rate to messages
• ✅ Active engagement (comments, responses)

Weekly Metrics

• ✅ 25 prospects → connected
• ✅ 10 connected → engaged in conversation
• ✅ 3 engaged → signed up / converted

Platform Health (Traffic Light System)

🟢 Green: >30% acceptance, >20% response, >3 signups/week 🟡 Yellow: 20-30% acceptance, 10-20% response, 1-2 signups/week 🔴 Red: <20% acceptance, <10% response, <1 signup/week

Goal: Stay in green with 20 minutes/week of your time.

Translation: You're getting $120k/year marketing manager results for:

• Cost: ~$20-50/month in AI credits (depending on model preferences)
• Time: 20 minutes/week of your attention
• Team size: 1 (you)

Contributing: Build Skills, Grow the Army

xswarm-buzz is powered by community-built Skills.

Create a Skill

npx xswarm-buzz
> "Create a new skill called 'reddit-engagement'"

Uses built-in skill-creator (from Anthropic). Walks you through building a skill with best practices.

Skill Structure

my-skill/
├── SKILL.md              ← Instructions + metadata (YAML frontmatter)
├── scripts/              ← Optional scripts
└── resources/            ← Optional resources

SKILL.md format: ```markdown

description: When Claude should use this skill (one clear sentence) tier: 2 preferred-models: [llama-3.2-local, claude-haiku] allowed-tools: [bash, browser] use-embeddings: false

Skill Name

Detailed instructions for Claude on how to use this skill...

### Share Your Skill

```bash
# Test it works
> "Use my new skill to [test scenario]"

# Contribute to community
git clone https://github.com/chadananda/xswarm-buzz
# Add your skill to skills/community/
# Submit PR

Maintainers review for:

• Quality (does it work?)
• Security (no credential leaks)
• Documentation (clear description)

Fetch Community Skills

> "Install the tweet-generator skill"
> "Show me available skills for email marketing"
> "Update all my skills to latest versions"

skill-fetcher handles downloading, updating, dependency resolution.

Roadmap: Skill marketplace, ratings, usage stats, featured skills.

Roadmap

Phase 1: Planning (Current)

Complete specification of all 13 planning documents:

• UX flows, conversation patterns, interface design
• Architecture, data schemas, authentication flows
• Campaign lifecycle, error handling, cost modeling
• Skill development guide, testing strategy

Status: In progress. No code yet. Planning first.

Phase 2: Implementation

Build the core system:

• npm wrapper with JSONL protocol
• Core skills (campaign-creator, message-generator, etc.)
• Multi-model orchestration
• Subagent delegation
• Context loading and learning system

Timeline: TBD after planning complete.

Phase 3: Beta

Invite developer-marketers to test:

• Real campaigns
• Feedback on UX
• Community skill contributions
• Model preference tuning

Timeline: TBD after planning complete.

Phase 4: Public Launch

• Stable API
• Comprehensive skill marketplace
• Documentation and tutorials
• Integration with xSwarm-Boss

Part of the xSwarm Ecosystem

xSwarm = Building the future where one developer = one complete company

The Family

🎯 xSwarm (npx xswarm)

• AI-driven project planning & coding
• Super detailed feature epoch planning
• Your AI development team

🎙️ xSwarm-Boss (Rust, voice-first)

• Personal assistant managing your projects, tasks, machines
• Calendar orchestration
• Talks to all your tools (including xswarm-buzz)
• Voice interface for hands-free control

📣 xswarm-buzz (npx xswarm-buzz) ← YOU ARE HERE

• AI marketing department
• Guerilla marketing automation
• Context-driven, conversational

The Philosophy

"Team of One" → You don't need to hire. You orchestrate AI agents.

"Be the team you want to lead" → AI agents as teammates, not tools.

Maximum capability. Zero overhead. → Build, plan, market, manage—all solo.

Based on "Army of One" → One developer, properly equipped with AI, can do what used to require a whole company.

Learn More

🌐 Website: xswarm.ai 💬 Discord: Join the community (coming soon) 🐙 GitHub: github.com/chadananda/xswarm-buzz

FAQ

"Is this just a chatbot wrapper around ChatGPT?"

No. This is:

• Multi-model orchestration (Claude, GPT, Gemini, local models)
• Cost optimization via subagent delegation
• Context-driven (reads your project files)
• Learning system (improves from outcomes)
• Skills-based extensibility (modular capabilities)
• JSONL protocol for machine control

It's an AI-native marketing automation system. The conversational interface is just how you interact with it.

"Can I use this if I'm not a developer?"

Technically yes, but it's built for developer-marketers who:

• Understand cd, npm, basic CLI
• Want to customize/extend via skills
• Appreciate the architecture (multi-model, cost optimization)
• Prefer conversation over dashboards

If you're non-technical, there are simpler tools. This is for people who want power and control.

"What about existing marketing tools (Hootsuite, Buffer, etc.)?"

Those are schedulers. You still do the creative work, strategy, personalization manually.

xswarm-buzz:

• Generates messages (AI, not templates)
• Plans strategy (researches market, builds campaigns)
• Learns patterns (A/B tests, optimizes automatically)
• Coordinates across platforms (unified intelligence)

Different category. This is your marketing manager, not your calendar.

"How much does it cost to run?"

AI Credits (main cost):

• Conservative: ~$20-30/month (using local models for Tier 2/3)
• Balanced: ~$50-75/month (mix of cloud and local)
• Premium: ~$100-150/month (all cloud, Tier 1 models)

Compare to:

• Marketing manager salary: $120,000/year ($10,000/month)
• Fractional marketer: $3,000-5,000/month
• Marketing agency: $5,000-15,000/month

Your time saved: 19h 40m/week = ~80 hours/month = $4,000-8,000 of your time (at $50-100/hr)

ROI is insane.

"What about privacy? I don't want my data sent to OpenAI."

Options for privacy:

1. Local models only: Set preferLocal: true, use Llama/Phi for everything
2. Hybrid: Use local for Tier 2/3 (data processing), cloud for Tier 1 (creative)
3. Self-hosted: Run your own Claude/GPT API proxy with data retention controls

All prospect data, interactions, auth tokens stored locally. Never sent to any AI unless needed for task execution. You control what data goes where.

"Can this get my LinkedIn account banned?"

Short answer: No, because YOU review and approve everything.

This isn't a spam bot. You're doing the marketing - AI just drafts faster.

Built-in safety:

• Human approval required - Nothing sends without your explicit review
• Respects rate limits (15 connections/day default, configurable)
• Human-like timing (randomized delays, not robotic)
• Natural language (AI-generated, not templates)
• Edit any message before sending

Your responsibility:

• Review messages before approving (you'll do this naturally)
• Don't be spammy (trust the defaults)
• Provide good context (product, value prop)
• Monitor acceptance rates (low rate = adjust strategy)

LinkedIn sees: You sending human-quality messages at reasonable volume with good engagement. Because that's exactly what you're doing - you're just using AI as your drafting assistant.

"What if I want to build my own skills?"

YES. That's the whole point.

Built-in skill-creator walks you through:

1. Defining when skill should trigger
2. Writing instructions for Claude
3. Adding optional scripts/resources
4. Testing with sample scenarios

Skill = folder with SKILL.md file. That's it.

Examples: competitor monitoring, email warmup tracking, meme generation, webinar registration workflows, referral program management, etc.

Build it. Use it. Share it (optional).

"Does this work for B2B? B2C? SaaS? E-commerce?"

B2B / SaaS (primary target):

• LinkedIn outreach ✅
• Personalized cold email ✅
• Community engagement (Reddit, HN, forums) ✅
• Content marketing (blog, social) ✅

B2C (works, different strategy):

• Facebook groups ✅
• Instagram/TikTok (community skills) ✅
• Reddit ✅
• Influencer outreach ✅

E-commerce (possible, less ideal):

• Paid ads (not automated by this tool)
• Influencer coordination ✅
• Social engagement ✅
• Email campaigns ✅

Best for: Startups, indie hackers, developer tools, SaaS, consulting, agencies.

Development & Testing

Testing the Interview Flow

Want to test the AI-powered interview without affecting your real projects?

Quick Test (Bash):

./test-interview.sh

This script:

1. Creates /tmp/testpromo-[random]/ directory
2. Copies your config/.env file (if exists)
3. Runs xswarm-buzz init in the test environment
4. Shows you the generated files

Manual Test:

# Create test environment
TEST_DIR="/tmp/testpromo-$(openssl rand -hex 4)"
mkdir -p "$TEST_DIR/config"

# Copy your .env (optional - or let init configure AI)
cp config/.env "$TEST_DIR/config/.env"

# Run init in test environment
cd "$TEST_DIR"
npx xswarm-buzz init

# Inspect results
ls -la
cat project-profile.json
cat campaigns/active/*.md

# Clean up when done
cd -
rm -rf "$TEST_DIR"

Using Environment Variables

For Claude Code users (recommended):

# No setup needed - init will detect your authenticated Claude session
npx xswarm-buzz init

For API key users:

# Export your API key before running
export ANTHROPIC_API_KEY="sk-ant-api03-..."
npx xswarm-buzz init
# Init will detect and offer to use your existing key

For Ollama users:

# Install and run Ollama first
brew install ollama
ollama pull llama3.2
# Init will auto-detect Ollama
npx xswarm-buzz init

Repository Structure

xswarm-buzz/
├── bin/cli.js                 # CLI entry point
├── src/
│   ├── commands/init.js       # AI-powered initialization
│   ├── modes/human.js         # Conversational interface
│   └── modes/machine.js       # JSONL protocol for automation
├── templates/                 # Project templates (populated during init)
│   ├── project-profile.json
│   ├── channels.json
│   ├── metrics.json
│   └── quick-reference.md
├── resources/
│   └── promotional-vectors/   # Platform-specific marketing strategies
│       ├── vector_twitter.md
│       ├── vector_linkedin.md
│       └── ... (8 total)
├── .claude/skills/
│   └── skill-creator/         # Anthropic's skill-creator skill (bundled)
└── test-interview.sh          # Test environment script

License

MIT License - Do whatever you want with this.

Let's Ship

Stop reading. Start marketing.

npx xswarm-buzz init

You're one developer. You're about to become an entire marketing department.

Team of One. Maximum capability. Zero overhead.

Welcome to xSwarm.

Built with: Claude Code, AI agents, too much coffee, and the conviction that one developer should be able to do everything.

Maintained by: @chadananda and the xSwarm community

Join us: xswarm.ai | GitHub | Discord (coming soon)