Skip to main content

​
Overview

Claude Code reads project instructions from CLAUDE.md in the repository root. Drevon generates this file with your full config plus Claude-specific features like allowedCommands.

​
Output File

CLAUDE.md

​
Special Features

​
Allowed Commands

Claude Code has a command execution permission system. Drevon automatically generates an allowedCommands section: 
## Allowed Commands

The following commands are pre-approved for use without user confirmation:

- `git`
- `npm`
- `npx`
- `node`
- `python3`
Configure in drevon.config.json: 
{
  "agents": {
    "claude": {
      "enabled": true,
      "allowedCommands": ["git", "npm", "npx", "node", "python3", "docker"]
    }
  }
}
Only add commands you trust Claude to run autonomously. Each listed command can be executed without confirmation prompts.

​
Enabling

drevon add-agent claude
When added via CLI, default allowedCommands (git, npm, npx, node, python3) are automatically configured.

​
Agent-Specific Settings

{
  "agents": {
    "claude": {
      "enabled": true,
      "allowedCommands": ["git", "npm", "npx", "node", "python3"],
      "extraInstructions": [
        "Always explain your reasoning before making changes",
        "Prefer small, focused commits"
      ]
    }
  }
}

​
Requirements

  • Claude Code installed (npm install -g @anthropic-ai/claude-code)
  • CLAUDE.md in the repository root (Drevon handles this)