Get the FREE Ultimate OpenClaw Setup Guide →

safe -manager

SafeMCP is an MCP tooling manager with simplicity and safety built in for your AI tools and Agents.

Installation
Run this command in your terminal to add the MCP server to Claude Code.
Run in terminal:
Command
claude mcp add --transport stdio openshieldai-safe-mcp-manager docker run -i safemcp-manager:latest \
  --env SAFEMCP_DEBUG="0" \
  --env SAFEMCP_APP_ENV="production" \
  --env SAFEMCP_HTTP_PORT="8080"

How to use

Safe MCP Manager provides a unified dashboard to manage MCP tools across AI applications (e.g., Claude, Cursor) with security-focused features. It orchestrates tool access, environment variables, and secure routing through a smart proxy, while sandboxed containers isolate each tool for safety. Use the dashboard to enable or disable individual tools per application, configure environment restrictions, and monitor tool status across Docker containers. The proxy integration ensures inputs and outputs are scanned for prompt injection risks before they reach the AI, helping maintain a secure MCP environment.

To start using, install the Docker-based Safe MCP Manager, launch the container, and access the web dashboard. From there you can connect your AI applications, assign available tools (such as context methods, file systems access, or web crawlers), and set per-tool permissions and tokens. Tools run in isolated containers, and the smart proxy routes requests to the correct MCP server, preserving security and auditability while enabling offline/local-first operation.

How to install

Prerequisites

  • Docker installed on your system (Docker Desktop for macOS/Windows or Docker Engine on Linux)
  • Administrative privileges to run containers

Installation steps

  1. Ensure Docker is running and updated:

    • macOS/Windows: open Docker Desktop
    • Linux: sudo systemctl start docker
  2. Pull or run the Safe MCP Manager image (as provided by Safe MCP):

    • docker pull safemcp-manager:latest
  3. Run Safe MCP Manager in a container:

    • docker run -d --name safemcp-manager -p 8080:8080 safemcp-manager:latest
  4. Open the web dashboard:

    • Navigate to http://localhost:8080 in your browser
    • Create your account and begin configuring connections to your AI applications (e.g., Claude, Cursor)
  5. Configure environment variables and tool permissions from the dashboard as needed. This setup is designed for offline/local-first usage with secure proxy routing.

Prerequisites note: This repository is for release management and usage analytics and does not contain the application source code. Obtain the official Safe MCP Manager image from the Safe MCP distribution channels (website or official registry).

Additional notes

Tips and common considerations:

  • The Safe MCP Manager relies on Docker for isolation; ensure your Docker daemon has enough resources (CPU/m memory) for running multiple tool containers.
  • Use per-tool environment variables to restrict access – for example, limit tokens, paths, or external endpoints per tool.
  • The Smart Proxy feature helps route requests securely; keep the proxy is up-to-date to benefit from prompt injection protection and threat detection.
  • If you encounter connectivity or tool activation issues, verify container permissions and ensure the host firewall allows the configured port (default 8080).
  • This repo focuses on release management and analytics; for tool configurations, refer to the Safe MCP Manager documentation on the official site.

Related MCP Servers

Sponsor this space

Reach thousands of developers