Hook CLI MCP

Advanced Event-Hook Automation Server for AI-Driven Business Process Remediation

Hook CLI MCP gives AI assistants the power to listen to real-time business events and execute trusted CLI commands inside hardened sandboxes. Connect your Stripe webhooks, Jira tickets, CI pipelines, and support queues directly to the AI.

Features

• Event-Driven Automation — Webhooks, cron, MQ triggers
• Security-by-Design — Docker/seccomp sandboxing, regex guards, secret masking
• hook_validate Dry-Run — Safe command testing before deployment
• AI Bottleneck Analysis — RAG pipeline over execution history
• Human-in-the-Loop — Slack/Teams approval for high-risk actions
• Full Observability — OpenTelemetry, Prometheus, Jaeger
• RBAC & Audit — Workspace-scoped roles, immutable audit trail

Quick Start

cp .env.example .env
docker compose up -d
curl http://localhost:8000/api/health

MCP Setup

pip install mcp
mcp dev .

Add to mcp.json for Claude Desktop / Cursor:

{
  "mcpServers": {
    "hookcli-mcp": {
      "command": "uv",
      "args": ["run", "hookcli-mcp"]
    }
  }
}

Development

python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -e ".[dev]"
pytest tests/ -v --asyncio-mode=auto
ruff check .

Architecture

AI Client (Claude) ──MCP──► Hook CLI MCP Server
                               ├── hook_validate (sandboxed dry-run)
                               ├── hook_create (webhook/cron/mq)
                               ├── bottleneck_analyze (RAG)
                               └── request_approval (Slack HITL)
                                        │
                               Docker Sandbox (seccomp, no-network, cap-drop)
                                        │
                               OTel ──► Jaeger / Prometheus

License

MIT