agentdock-hook

Claude Code hook runner for AgentDock — streams every tool call to your dashboard and blocks on file writes until a human approves them.

Install

npm install -g agentdock-hook

Setup

1. Get your project API key from AgentDock → Project → Settings.

2. Set environment variables (add to your shell profile or .env):

AGENTDOCK_API_KEY=ah_live_...
AGENTDOCK_BASE_URL=https://www.getagentdock.com
AGENTDOCK_REQUIRE_APPROVAL=true

3. Add hooks to .claude/settings.json in your project:

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Write|Edit|MultiEdit|Bash",
        "hooks": [{ "type": "command", "command": "npx agentdock-hook pre", "timeout": 30 }]
      }
    ],
    "PostToolUse": [
      {
        "matcher": ".*",
        "hooks": [{ "type": "command", "command": "npx agentdock-hook post" }]
      }
    ]
  }
}

4. Run Claude Code. Every session, tool call, and file change will appear in your AgentDock dashboard in real time.

How it works

• Pre-hook (blocking): fires before every Write/Edit/Bash. Posts the event to AgentDock, then polls for approval. If approved, exits 0. If rejected, exits 2 — Claude Code shows the rejection reason and stops.
• Post-hook (non-blocking): fires after every tool call. Records the result and token usage.

Environment variables

| Variable | Default | Description | |----------|---------|-------------| | AGENTDOCK_API_KEY | — | Required. Your project API key. | | AGENTDOCK_BASE_URL | http://localhost:3000 | AgentDock instance URL. | | AGENTDOCK_REQUIRE_APPROVAL | true | Set to false to record events without blocking. | | AGENTDOCK_TIMEOUT_ACTION | approve | What to do if approval poll times out: approve or reject. |

Requirements

Node.js ≥ 18