AI Employee Vault
daniyalazhar123/AI-Employee-Vault
Agente IA autônomo multi-canal com Odoo, Gmail, redes sociais e aprovação humana
Ver no GitHubMCP server que implementa um agente autônomo multi-canal para gerenciar e-mails, Odoo, redes sociais e WhatsApp via watchers integrados. Processa tarefas automaticamente usando Claude Code como engine de IA, com aprovação humana opcional e recuperação de erros via circuit breaker.
Instalação
npx convez-ai daniyalazhar123/AI-Employee-Vault1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Automatizar fluxo de e-mails, leads Odoo e redes sociais simultâneamente
- Executar tarefas repetitivas com aprovação humana (human-in-the-loop)
- Monitorar múltiplos canais e consolidar ações em vault de conhecimento
Veredito

Projeto ambicioso de hackathon com arquitetura sólida (watchers → vault → MCP), mas depende de muita configuração manual (Docker, APIs, secrets) e está em estágio Gold em desenvolvimento. Viável para quem quer automação end-to-end, mas requer expertise em DevOps e integração de APIs.
README
A tradução para o português deste README ainda está na fila.
Personal AI Employee — Hackathon Project
Hackathon: Personal AI Employee Hackathon 0
Stack: Python 3.13+ / Obsidian / Claude Code / Docker
Vault: D:\Desktop4\Obsidian Vault
Secrets: C:\Users\%USERNAME%\.ai_employee\secrets\
GitHub: https://github.com/daniyalazhar123/AI-Employee-Vault
Quick Status
| Tier | Status | Verified |
|---|---|---|
| 🥉 Bronze | ✅ Complete | Yes |
| 🥈 Silver | ✅ Complete | Yes |
| 🥇 Gold | ✅ Complete | June 10, 2026 |
| 💎 Platinum | ❌ Not Started | — |
What This Does
- Monitors Gmail, WhatsApp, LinkedIn, Odoo automatically
- Processes tasks using Claude Code as AI engine
- Creates invoices in Odoo 17 (Docker)
- Sends real emails via Gmail API
- Posts to LinkedIn / Facebook / Instagram / Twitter
- Human-in-the-loop approval for sensitive actions
- Ralph Wiggum loop for autonomous task completion
- Error recovery with circuit breaker and dead letter queue
- Audit logging for all actions
Architecture
Email ─┐
WhatsApp ─┤ ┌─────────────┐
Social ─┼── Watchers ─────→ │ Obsidian │ ←── Claude Code (AI Engine)
Odoo ─┘ │ │ Vault │ │
│ │ (File System)│ │
│ └─────────────┘ │
│ │ │
▼ ▼ ▼
┌──────────┐ ┌──────────┐ ┌──────────┐
│ mcp_email │ │ mcp_odoo │ │mcp_social│
└──────────┘ └──────────┘ └──────────┘
│ │ │
▼ ▼ ▼
Gmail API Odoo 17 (Docker) LinkedIn/FB/IG/Twitter
Flow: Watchers detect events → write .md files to vault folders → Claude Code reads tasks → executes via MCP servers.
Setup
Prerequisites
- Docker Desktop (v29+)
- Python 3.13+
- Node.js v24+ (for Claude Code)
- Claude Code CLI (
npm install -g @anthropic-ai/claude-code) / OpenRouter (DeepSeek V3) - Git
Secrets Setup
All credentials stored outside vault at C:\Users\%USERNAME%\.ai_employee\secrets\:
DRY_RUN=false
REQUIRE_APPROVAL=false
ODOO_URL=http://localhost:8069
ODOO_DB=odoo
ODOO_USERNAME=admin
ODOO_PASSWORD=admin
Files in secrets directory:
.env— Merged configurationcredentials.json— Gmail API OAuth credentialstoken.pickle— Gmail OAuth tokenlinkedin_session.json— LinkedIn session cookies
Start Odoo
docker compose up -d
# Odoo 17 at http://localhost:8069
# PostgreSQL 15 at localhost:5432
# Default: admin / admin
Run
# Start all watchers
python orchestrator.py
# Or individual watchers
python watchers/gmail_watcher.py
python watchers/whatsapp_watcher.py
python watchers/social_watcher.py
python watchers/odoo_lead_watcher.py
# MCP servers (manual)
python mcp_email.py --action list
python mcp_odoo.py --action get_partners
python mcp_social.py --action status
# Ralph Wiggum loop (persistent task execution)
python ralph_loop.py "Process all files in Needs_Action"
Vault Structure
Obsidian Vault/
├── .claude/
│ ├── skills/ # 8 agent skills
│ └── README.md
│
├── watchers/ # Perception layer (5 watchers)
│ ├── gmail_watcher.py
│ ├── whatsapp_watcher.py
│ ├── social_watcher.py
│ ├── office_watcher.py
│ └── odoo_lead_watcher.py
│
├── Needs_Action/ # Incoming tasks
├── Pending_Approval/ # Drafts awaiting human review
├── Done/ # Completed tasks
├── Drafts/ # Draft content
├── Social_Drafts/ # Social media drafts
├── Social_Summaries/ # Social media reports
├── Logs/ # Batch processing logs
├── Dead_Letter_Queue/ # Failed items
├── Approved/ # Approved action files
│
├── mcp_email.py # Email MCP server
├── mcp_odoo.py # Odoo MCP server
├── mcp_social.py # Social media MCP server
├── orchestrator.py # Master orchestrator
├── ralph_loop.py # Persistent task executor
├── batch_processor.py # Backlog processor
├── secrets_config.py # Centralized secrets loader
├── docker-compose.yml # Odoo 17 + PostgreSQL 15
├── STATUS.md # Current real status
├── Dashboard.md # Live dashboard (counts)
└── README.md # This file
Tier Documentation
| Tier | Document | Status |
|---|---|---|
| 🥉 Bronze | docs/BRONZE_TIER.md | ✅ Complete |
| 🥈 Silver | docs/SILVER_TIER.md | ✅ Complete |
| 🥇 Gold | docs/GOLD_TIER.md | ✅ Verified June 10, 2026 |
| 💎 Platinum | docs/PLATINUM_TIER.md | ❌ Not Started |
Security
Zero credentials in the vault. All secrets are stored at C:\Users\%USERNAME%\.ai_employee\secrets\ and loaded via secrets_config.py. The vault is safe to push to public GitHub.
Files excluded via .gitignore:
.env,*.pickle,credentials.json,token.json,linkedin_session.json__pycache__/,*.pycLogs/,In_Progress/linkedin_browser_data/,linkedin_browser_profile/
GitHub
https://github.com/daniyalazhar123/AI-Employee-Vault
Last updated: June 10, 2026 — Gold Tier Verified