Claude Code Integration

Claude Code is Anthropic’s agentic command-line tool. By routing Claude Code’s tools through dmint-mcp, you ensure that autonomous file edits, database mutations, and external API requests follow deterministic security boundaries.

1. Install the Policy Skill

Install the official dmint-policy-manager skill into Claude Code’s global configuration:
This gives Claude Code full understanding of Dmint’s policy syntax, failure modes, and approval retries.

2. Configure Claude Code with dmint-mcp

Configure Claude Code to route downstream tools through the dmint-mcp gateway. Add the gateway to your Claude Code configuration:
Alternatively, configure ~/.claude/config.json:
~/.claude/config.json

3. Run Claude Code

Start your Claude Code session as usual:
When Claude Code requests a tool call:
  1. ALLOW tools (e.g. reading logs, running tests) execute immediately.
  2. DENY tools (e.g. dropping production databases, deleting root files) fail closed immediately with a structured security error.
  3. APPROVAL_REQUIRED tools halt and wait for human review on your local dashboard.

4. Approve via CLI or Dashboard

Run the approval dashboard alongside your terminal session:
Review the request at http://127.0.0.1:8080, inspect the parameters, and click Approve. Claude Code resumes execution immediately upon approval.