AgentSentinel¶
Auto-approve tool requests for AI coding CLIs running in tmux.
AgentSentinel monitors tmux panes for tool approval prompts from AI coding assistants and automatically responds with y to approve them. Designed for macOS with iTerm2 and tmux.
Supported AI CLIs¶
| CLI | Prompt Format |
|---|---|
| AWS Kiro CLI | tool use read requires approval, press 'y' to approve |
| Claude Code | Allow? (Y/n) |
| Codex CLI | Proceed? (Y/n) |
| Gemini CLI | (Y/n) prompts |
Key Features¶
- Automatic Approval - Detects and approves tool requests without manual intervention
- Multi-Agent Support - Handles multiple concurrent AI agents across different tmux panes
- Kiro Multi-Subagent TUI - Special handling for Kiro's navigable approval list
- Safety First - Blocks dangerous commands like
rm -rf,sudo,curl|bash - Configurable - Custom patterns, notifications, and statistics tracking
- Dry Run Mode - Test detection without sending approvals
How It Works¶
┌─────────────────────────────────────┐
│ tmux pane 1: kiro │
│ │
│ > tool use read requires approval, │
│ press 'y' to approve │ ← AgentSentinel detects this
│ │
├─────────────────────────────────────┤
│ tmux pane 2: agentsentinel watch -v │
│ │
│ INFO prompt detected pane=%1 │
│ INFO approved pane=%1 │ ← Sends 'y' to pane 1
└─────────────────────────────────────┘
Quick Links¶
- Installation - Get AgentSentinel running
- Quick Start - Step-by-step guide
- Commands - CLI reference
- Configuration - YAML config options
- Pattern Reference - Detection and danger patterns
License¶
MIT