cco-mcp

🥈Silver

Real-time audit and approval system for Claude Code tool calls. Operations teams gain visibility and control over AI agent actions. Integrates with Claude Code to monitor and approve tool calls via a web dashboard.

1940Updated 2mo ago

Intermediate30min to implementautomation

Saves ~75 min per use

Quick InstallView Source

git clone https://github.com/toolprint/cco-mcp.git

Works with:

Claude

Overview

About This Skill

CCO-MCP provides a security layer between Claude Code and your system, enabling real-time monitoring and control of AI tool calls. It acts as a firewall for AI actions, allowing you to review sensitive operations before execution while auto-approving safe ones based on configurable rules. The system features a live web dashboard for instant visibility, detailed audit logs, and smart approval rules that match specific tools, agents, and patterns. Operations teams can scale background Claude Code instances without manual intervention by setting up rules for routine operations. Installation is straightforward via Docker Compose with integration into Claude Code's MCP server configuration.

How to Use

Clone the repository and run `docker-compose up`. Install CCO-MCP in Claude Code using `claude mcp add -s user -t http cco-mcp http://localhost:8660/mcp`. Configure approval rules through the web dashboard at http://localhost:8660. For non-interactive mode, use the `--permission-prompt-tool mcp__cco-mcp__approval_prompt` flag.

Use Cases

Auto-approve safe file read operations while requiring manual approval for bash commands

Monitor and control multiple background Claude Code agents from a centralized dashboard

Scale non-interactive Claude Code instances with rule-based auto-approval for routine operations

Audit all AI tool calls with full context including agent identity and approval status

Setup & Installation

Quick Install

No install command available. Check the GitHub repository for manual installation instructions.

Alternative Install (Git Clone)

git clone https://github.com/toolprint/cco-mcp

Requirements

Claude Code or compatible AI agent
Works with: Claude

Quick Start Guide

Install the Skill

Copy the install command above and run it in your terminal.

Open Your AI Agent

Launch Claude Code, Cursor, or your preferred AI coding agent.

Try It Out

Use the prompt template or examples below to test the skill.

Customize

Adapt the skill to your specific use case and workflow.

Usage Examples

Prompt Template

Act as a real-time audit and approval system for Claude Code tool calls. Monitor all tool calls made by AI agents in [COMPANY] within the [INDUSTRY] sector. Provide a dashboard view of pending approvals, approved actions, and rejected actions. Allow operations teams to review and approve or reject tool calls in real-time. Include details such as [DATA] for each tool call.

Example Output

# Real-Time Audit and Approval Dashboard

## Pending Approvals

1. **Tool Call ID: TC-20240521-001**
   - **Agent:** Inventory Management Bot
   - **Action:** Update inventory levels
   - **Details:** Adjust stock quantities for SKU-12345 based on recent sales data.
   - **Status:** Pending Approval

2. **Tool Call ID: TC-20240521-002**
   - **Agent:** Customer Support Bot
   - **Action:** Send follow-up email
   - **Details:** Send a follow-up email to customer [EMAIL] regarding their recent purchase.
   - **Status:** Pending Approval

## Approved Actions

1. **Tool Call ID: TC-20240521-003**
   - **Agent:** Billing Bot
   - **Action:** Generate invoice
   - **Details:** Generate invoice INV-20240521-001 for customer [CUSTOMER_ID].
   - **Status:** Approved by [USERNAME]

## Rejected Actions

1. **Tool Call ID: TC-20240521-004**
   - **Agent:** Marketing Bot
   - **Action:** Send promotional email
   - **Details:** Send promotional email to all customers in the [INDUSTRY] sector.
   - **Status:** Rejected by [USERNAME]
   - **Reason:** Email content not reviewed for compliance.

## Summary Statistics

- **Total Pending Approvals:** 2
- **Total Approved Actions:** 1
- **Total Rejected Actions:** 1
- **Approval Rate:** 50%
- **Average Approval Time:** 2 minutes 30 seconds