# Dmcps

> Dmcps is an open-source MCP server for security. Sandboxed MCP server giving AI coding agents shell and file access in Docker or Cloud Run. It is available on CLI, API and self-hosted.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, API, self-hosted
- **Repository:** https://github.com/rmaapk/dmcps
- **MCP Registry name:** io.github.thealidev/dmcps
- **Categories:** [security](https://ai-powered.com/categories/security)
- **Listing updated:** 2026-11-18

## About

The README says DMCPS runs an MCP server that lets AI coding agents such as Claude Desktop, Cursor or Gemini run terminal commands and read or write files, with execution kept inside a disposable container rather than on the host machine. Access is limited to directories an owner whitelists through a web dashboard, and clients connect with an auto-generated Bearer token.

The project is a Node.js/Express backend daemon. It is documented for deployment to Google Cloud Run through Cloud Build using the included cloudbuild.yaml, and for local testing with docker compose. The dashboard login uses Product Hunt PKCE OAuth, and the MCP tools include a Product Hunt GraphQL query tool alongside file and shell tools.

The code is published under the BSL-1.1 licence, which the README says allows inspection and deployment to your own GCP environment while restricting commercial use. It suits developers who want to give coding agents terminal access without exposing their own machines.

## Features

- **Whitelisted directory access:** The AI can only read, write or run commands inside folders the owner has explicitly allowed in the dashboard.
- **Disposable container sandbox:** Commands run inside a non-root Alpine container with Linux capabilities dropped, so the host machine is not touched.
- **Bearer token authentication:** Each MCP client connects using an automatically generated Bearer token that controls which agents can call tools.
- **Cloud Run deployment:** A cloudbuild.yaml file builds the container and deploys it as a Google Cloud Run service.
- **Product Hunt OAuth login:** The dashboard uses PKCE OAuth 2.0 so only Product Hunt makers can sign in and manage sandboxes.
- **Preinstalled developer tools:** Git, python3, curl, bash, make and jq are included in the image so agents can start working immediately.

## Good for

- Letting a coding agent run builds and tests without access to the developer's own machine
- Giving an AI assistant controlled read and write access to a project folder
- Hosting a shared sandbox for agents on Google Cloud Run
- Querying Product Hunt analytics from an AI client through the ph_graphql_query tool

## Tools

- `ph_graphql_query`: Run analytics queries against the Product Hunt API.
- `read_file`: Read text from a file.
- `write_file`: Write content to a file.
- `list_directory`: List all files in a folder.
- `run_shell_command`: Execute terminal commands within the isolated workspace.

## FAQ

### Which tools does the MCP server expose?

The README lists ph_graphql_query, read_file, write_file, list_directory and run_shell_command.

### How is the AI prevented from accessing other files?

The README says it cannot read, write or execute outside directories whitelisted in the web dashboard, and that directory traversal such as ../ is blocked.

### What licence is the code under?

The README states the codebase is under BSL-1.1, which allows inspection and deployment to your own GCP environment while restricting commercial and enterprise use.

### Can it be run locally?

Yes, the README gives steps to run the stack locally or on AWS with docker compose up -d --build.

### Is Dmcps free?

Yes. Dmcps is open source, so you can use and self-host it for free.

### What are the best alternatives to Dmcps?

The closest alternatives to Dmcps are ClosedMCP, Linux MCP daemon (mcpd), mcpcut, Headmesh Agent Sandbox and Diogenes Agent Guardrails & Sandbox MCP.

## Where this comes from

- Description: written from the project's README and its MCP Registry entry.

## Alternatives

- [ClosedMCP](https://ai-powered.com/mcp/closedmcp): ClosedMCP is a free MCP server for developer tools.
- [Linux MCP daemon (mcpd)](https://ai-powered.com/mcp/linux-mcp-daemon-mcpd): Linux MCP daemon (mcpd) is an open-source MCP server for cloud and infrastructure.
- [mcpcut](https://ai-powered.com/mcp/mcpcut): mcpcut is an open-source MCP server for security.
- [Headmesh Agent Sandbox](https://ai-powered.com/mcp/headmesh-agent-sandbox): Headmesh Agent Sandbox is an open-source MCP server for developer tools.
- [Diogenes Agent Guardrails & Sandbox MCP](https://ai-powered.com/mcp/diogenes-agent-guardrails-and-sandbox-mcp): Diogenes Agent Guardrails & Sandbox MCP is a freemium MCP server for security.
- [Whiskers](https://ai-powered.com/mcp/whiskers): Whiskers is an open-source MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/dmcps
