BE Agent Service

Multi-agent autonomous service for software development and marketing automation with closed-loop WhatsApp/Telegram integration.

Overview

Three agent teams (23 agents total):

Engineering (9 specialists): Backend, Frontend, Infrastructure, Verification, DB Architect, UX Designer, Documentation Expert, Agent Levelup, Orchestrator
Marketing (10 Marvel agents): Jarvis, Shuri, Fury, Vision, Loki, Quill, Wanda, Pepper, Friday, Wong
Management (4 executives): CEO, CPO/CTO, CMO/CSO, HR Agent Lead

Human-AI Interface:

WhatsApp/Telegram → OpenClaw → Shared Workspace → Agent Service → Notifications back to you
Dashboard: http://localhost:3030

Automation:

Nightly: 10:30 PM learnings extraction, 11:00 PM auto-implementation
On-demand: Trigger via Telegram or dashboard
Notifications: Session complete, morning briefing (8 AM), weekly review (Monday 8 AM)

Unified Dashboard (port 3030)

Single entry point at http://localhost:3030:

URL	Content
`/`	React workspace UI (repo selector, workspace, plans, agents, logs)

How to run:

yarn start          # Build + start (single server on 3030)
yarn dev            # Build + dev with hot reload

Architecture: One Express server (apps/server) on port 3030 serves API + static. See docs/ARCHITECTURE.md.

Documentation

All docs live in docs/. Start with docs/README.md for the full index.

Need	Doc
Closed-loop integration	docs/CLOSED_LOOP_INTEGRATION.md ⭐
Workspace setup	docs/WORKSPACE.md
OpenClaw setup	config/openclaw/SIMPLE_SETUP.md
Quick commands	docs/QUICK_REFERENCE.md
Mac mini setup	docs/FRESH_MAC_MINI_SETUP.md
Mac mini recovery	docs/MAC_MINI_RECOVERY.md
Architecture	docs/ARCHITECTURE.md
Data flow	docs/DATA_FLOW.md
Database access	docs/DATABASE_ACCESS.md
API reference	docs/API_ENDPOINTS.md
Dashboard migration	docs/DASHBOARD_MIGRATION.md

Repository Structure

be-agents-service/
├── agents/
│   ├── marketing/                   # Marketing agent team (10 Marvel characters)
│   │   ├── jarvis-orchestrator.sh   # Squad lead
│   │   ├── vision-seo-analyst.sh    # SEO specialist
│   │   └── ... (8 more agents)
│   ├── backend-specialist.sh        # Engineering: Database, GraphQL, resolvers
│   ├── frontend-specialist.sh       # Engineering: UI, codegen, operations
│   ├── infrastructure-specialist.sh # Engineering: Packages, configs, docs
│   └── verification-specialist.sh   # Engineering: Quality gate
├── scripts/
│   ├── orchestrator.sh              # Multi-repo coordinator
│   ├── daily-compound-review.sh     # 10:30 PM - Extract learnings
│   ├── auto-compound.sh             # 11:00 PM - Auto-implement
│   ├── loop.sh                      # Task execution loop
│   ├── analyze-report.sh            # Parse priorities report
│   ├── check-status.sh              # Status monitoring
│   └── test-safety.sh               # Safety mechanism tests
├── lib/
│   ├── state-manager.sh             # JSON state coordination
│   ├── feedback-schema.json         # Agent communication schema
│   └── parallel-executor.sh         # Parallel agent spawning
├── apps/
│   ├── server/                      # Unified server (port 3030): API + static in public/
│   └── dashboard/                   # React workspace UI (build → server/public)
├── scripts/
│   └── start-dashboard.sh           # LaunchD: yarn workspace server start
├── launchd/
│   ├── com.appcaire.auto-compound.plist
│   ├── com.appcaire.caffeinate.plist
│   ├── com.appcaire.daily-compound-review.plist
│   └── com.appcaire.dashboard.plist
├── .compound-state/                 # Agent session states (JSON)
├── data/                            # SQLite database
├── docs/                            # All documentation
└── config/repos.yaml                # Multi-repo config

Installation

1. Clone Repository

cd ~/HomeCare
git clone https://github.com/bear75/be-agents-service.git
cd be-agents-service

2. Configure LaunchAgents

# Copy plist files to LaunchAgents directory
cp launchd/*.plist ~/Library/LaunchAgents/

# Load the agents
launchctl load ~/Library/LaunchAgents/com.appcaire.daily-compound-review.plist
launchctl load ~/Library/LaunchAgents/com.appcaire.auto-compound.plist
launchctl load ~/Library/LaunchAgents/com.appcaire.caffeinate.plist

3. Configure Paths

Edit plist files to update paths:

# Update WorkingDirectory and ProgramArguments paths
vim ~/Library/LaunchAgents/com.appcaire.*.plist

4. Test Installation

# Run daily review manually
./scripts/daily-compound-review.sh

# Run auto-compound manually
./scripts/auto-compound.sh

Schedules

Script	Schedule	Purpose
daily-compound-review.sh	10:30 PM daily	Extract learnings from Claude threads
auto-compound.sh	11:00 PM daily	Auto-implement priority #1 task
caffeinate	Always	Keep Mac awake for automation

Safety Mechanisms

1. Branch Check

✅ Must be on main branch before starting

2. Auto-Stash

✅ Automatically stashes uncommitted work

3. PR Creation

✅ Creates PR instead of direct main push

4. Safety Commit

✅ Auto-commits if Claude forgets

5. Error Handling

✅ Stops on errors, logs failures

Dashboard

Real-time monitoring at http://localhost:3030

# Start dashboard
cd ~/HomeCare/be-agents-service
./dashboard/start.sh

# Auto-starts on boot via LaunchAgent

Features:

Live session monitoring (3s refresh)
Agent status tracking
Session logs viewer
System statistics
Command center (all commands in one place)
Documentation browser

Usage

Engineering Agents (Automatic)

# Run daily review
cd ~/HomeCare/beta-appcaire
../be-agents-service/scripts/daily-compound-review.sh

# Run auto-compound
cd ~/HomeCare/beta-appcaire
../be-agents-service/scripts/auto-compound.sh

# Run orchestrator manually
cd ~/HomeCare/be-agents-service
./scripts/orchestrator.sh \
  ~/HomeCare/beta-appcaire \
  ~/HomeCare/beta-appcaire/reports/priorities-2026-02-07.md \
  ~/HomeCare/beta-appcaire/tasks/prd.json \
  feature/test-branch

# Check status
../be-agent-service/scripts/check-status.sh

# Test safety mechanisms
../be-agent-service/scripts/test-safety.sh

Marketing Agents (Manual)

# Create marketing priority
cat > ~/HomeCare/beta-appcaire/reports/marketing-blog.md <<EOF
# Priority: SEO Blog Post
**Description:** Create SEO-optimized blog about employee scheduling
**Expected outcome:**
- Keyword research
- Blog post written (1500+ words)
- Header image created
- Published to website
- Promoted on social media
EOF

# Run Jarvis marketing orchestrator
cd ~/HomeCare/be-agents-service
./agents/marketing/jarvis-orchestrator.sh \
  ~/HomeCare/beta-appcaire \
  ~/HomeCare/beta-appcaire/reports/marketing-blog.md \
  ~/HomeCare/beta-appcaire/tasks/marketing-prd.json \
  feature/blog-post-scheduling

# Run individual marketing agent
./agents/marketing/vision-seo-analyst.sh \
  "session-test-$(date +%s)" \
  ~/HomeCare/beta-appcaire \
  ~/HomeCare/beta-appcaire/reports/marketing-blog.md

Scheduled Execution

LaunchAgents handle automatic execution:

# Check if agents are loaded
launchctl list | grep appcaire

# View logs
tail -f ~/Library/Logs/appcaire-compound.log
tail -f ~/Library/Logs/appcaire-daily-review.log

# Manually trigger
launchctl start com.appcaire.auto-compound
launchctl start com.appcaire.daily-compound-review

Configuration

Repository Location

The service expects beta-appcaire at:

~/HomeCare/beta-appcaire

Update plist files if location differs.

Priority Reports

Daily priorities should be in:

~/HomeCare/beta-appcaire/reports/priorities-YYYY-MM-DD.md

Format:

# Priority 1

**Description:** Clear task description
**Expected outcome:** What should be achieved
**Files:** Relevant file paths

Logs

Logs are stored in:

~/Library/Logs/appcaire-compound.log
~/Library/Logs/appcaire-daily-review.log

View recent logs:

tail -f ~/Library/Logs/appcaire-*.log

Troubleshooting

Agent Not Running

# Check if loaded
launchctl list | grep appcaire

# Reload if needed
launchctl unload ~/Library/LaunchAgents/com.appcaire.auto-compound.plist
launchctl load ~/Library/LaunchAgents/com.appcaire.auto-compound.plist

Wrong Branch Error

# Ensure on main branch
cd ~/HomeCare/beta-appcaire
git checkout main
git pull origin main

Permission Errors

# Make scripts executable
chmod +x ~/HomeCare/be-agents-service/scripts/*.sh

Development

Testing Scripts

# Test daily review (dry run)
./scripts/daily-compound-review.sh --dry-run

# Test auto-compound (dry run)
./scripts/auto-compound.sh --dry-run

# Run safety tests
./scripts/test-safety.sh

Modifying Schedules

Edit plist files:

<key>StartCalendarInterval</key>
<dict>
    <key>Hour</key>
    <integer>22</integer>  <!-- 10 PM -->
    <key>Minute</key>
    <integer>30</integer>
</dict>

Then reload:

launchctl unload ~/Library/LaunchAgents/com.appcaire.auto-compound.plist
launchctl load ~/Library/LaunchAgents/com.appcaire.auto-compound.plist

Architecture

┌─────────────────────────────────────────┐
│  Mac Mini (Agent Host)                  │
│                                         │
│  LaunchAgent (10:30 PM)                │
│         ↓                               │
│  daily-compound-review.sh              │
│         ↓                               │
│  Reviews Claude threads                │
│  Updates CLAUDE.md files               │
│  Commits to main                        │
│                                         │
│  LaunchAgent (11:00 PM)                │
│         ↓                               │
│  auto-compound.sh                      │
│         ↓                               │
│  analyze-report.sh                     │
│         ↓                               │
│  loop.sh (Claude Code API)            │
│         ↓                               │
│  Creates feature branch                │
│  Implements task                        │
│  Creates PR                             │
│                                         │
└─────────────────────────────────────────┘
                 ↓
         GitHub Pull Request
                 ↓
         Human Review & Merge

Related Repositories

beta-appcaire: Main application repository
- https://github.com/[org]/beta-appcaire

License

Private - Internal use only

Contact

Issues: https://github.com/bear75/be-agents-service/issues

Name		Name	Last commit message	Last commit date
Latest commit History 46 Commits
agents		agents
apps		apps
config		config
dashboard		dashboard
data		data
docs-archive		docs-archive
docs		docs
launchd		launchd
lib		lib
scripts		scripts
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
DATABASE_ACCESS.md		DATABASE_ACCESS.md
DATA_FLOW.md		DATA_FLOW.md
README.md		README.md
SAFETY.md		SAFETY.md
gamification-schema.sql		gamification-schema.sql
package.json		package.json
reset-gamification.sql		reset-gamification.sql
schema.sql		schema.sql
seed.sql		seed.sql
test-task-completion.sql		test-task-completion.sql
test-xp-earning.js		test-xp-earning.js
yarn.lock		yarn.lock

bear75/be-agents-service

Folders and files

Latest commit

History

Repository files navigation