DMCPS MCP server
Commit Nov 1, 2026Provides a secure, sandboxed Docker environment on Google Cloud Run for AI agents to execute shell commands and manage files safely.
Also known as RMAAPK/dmcps
- Stars
- 1
- Tools
- 5
- Downloads, 30 days
- —
- Latest version
- 1.0.0
- Last commit
- Nov 1, 2026
- License
- —
What is DMCPS MCP?
DMCPS provides a secure, sandboxed Docker environment on Google Cloud Run for AI agents to execute shell commands and manage files safely. It exposes 5 tools, such as ph_graphql_query, read_file, write_file.
Summary generated from the project's README
About DMCPS MCP server
DMCPS provides a sandboxed Docker environment hosted on Google Cloud Run to allow AI agents to execute shell commands and manage files safely. This server is designed for developers using clients like Claude Desktop, Cursor, or Gemini who need to grant an agent terminal access without risking the host machine. It uses a Node.js and Express backend to manage a disposable containerized workspace where the agent operates as a root user within an isolated Alpine Linux environment.
Users deploy the service via Google Cloud Build using the provided cloudbuild.yaml configuration. Security is managed through a web dashboard that uses Product Hunt PKCE OAuth 2.0 for authentication and allows administrators to whitelist specific directories. To connect an agent, you use auto-generated Bearer tokens. The environment comes pre-loaded with essential tools including git, python3, curl, bash, make, and jq, ensuring the agent can immediately perform technical tasks within the restricted workspace.
What you can do with it
- 1Execute terminal commands safelyAn AI agent can run shell commands inside an isolated container to prevent any risk to your local computer.
- 2Manage files in restricted directoriesYou can instruct an AI to read or write files only within specific folders that you have approved via a dashboard.
- 3Perform Product Hunt data analysisA developer can ask an AI to run complex analytics queries against the Product Hunt API using built-in tools.
Written from the project's README. Every use case is checked against a quote from it.
DMCPS MCP tools 5
Detected from the READMEph_graphql_queryread_filewrite_filelist_directoryrun_shell_command
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablermaapk/dmcps is public and reachablegithub.com/rmaapk/dmcps ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)No licenseFinding
- Commit or release in the last yearLast commit or release 2026-11-01github.com/rmaapk/dmcps ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model5 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHub🛡️ DMCPS (Docker Model Context Protocol Secured)
DMCPS V3 is the ultimate enterprise-grade sandbox for AI coding agents, completely re-engineered to run natively on Google Cloud Run. By utilizing Google's serverless container infrastructure, you can give AI agents (like Claude Desktop, Cursor, or Gemini) full terminal and filesystem execution without ever risking your host machine.
🌟 Hackathon Exclusive Features
- Native Google Cloud Run Architecture: Zero-friction automated deployments via
cloudbuild.yaml. Designed from the ground up to utilize Google's serverless containers. - Product Hunt PKCE OAuth Flow: The web dashboard is secured using Product Hunt's official API. Authenticate your makers and manage your AI sandbox instantly using your Product Hunt account.
- Enterprise Open Source (BSL-1.1): Codebase is fully open for developers to inspect the kernel-level protections and deploy to their own GCP environment, while legally protecting commercial/enterprise use.
- Kernel-Level Locks: Non-root Alpine execution with dropped Linux capabilities (
cap_drop: ALL). - Absolute Root Illusion: The AI is assigned a direct
rootuser within the disposable container, tricking it into feeling absolute freedom without any "Permission denied" frustrations, keeping it from getting stuck in looping errors.
🚀 Getting Started on Google Cloud
Deploying DMCPS V3 is incredibly simple using Google Cloud Build.
- Navigate to Cloud Build Triggers in your Google Cloud Console.
- Click Connect Repository and link this GitHub repository.
- Set the trigger to watch the
v3-betabranch. - Point the configuration to the included
/cloudbuild.yamlfile. - Click Run.
Google Cloud Build will automatically build the container, push it to your Container Registry, and deploy it as a highly scalable Google Cloud Run service!
🔒 Security & Dashboard Whitelisting
This is built as a robust Node.js/Express backend daemon, featuring a "military-grade" secured dashboard to strictly manage which directories the AI is allowed to touch.
- Strict Whitelisting: The AI cannot read, write, or execute commands outside of directories explicitly whitelisted via the web dashboard. (Directory traversal attempts like
../are mathematically blocked). - Product Hunt Integration: Secure login restricted to Product Hunt makers using PKCE OAuth 2.0 flow.
- Auto-Generated API Keys: Connect to your MCP server using a dynamically generated Bearer token to ensure only authorized agents can execute tools on your server.
- Pre-installed AI Toolkit: Foundational tools (
git,python3,curl,bash,make,jq) are pre-baked into the image so the AI is immediately ready to work.
🛠️ MCP Tools Exposed to the AI
Once authenticated and restricted to a whitelisted folder, the AI has access to:
ph_graphql_query- Run deep 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 strictly within the isolated workspace.
🧪 Local Testing
You can easily test the entire stack locally or on an AWS instance before pushing to Google Cloud Run:
git clone -b v3-beta https://github.com/RMAAPK/dmcps.git
cd dmcps
git checkout v3-beta
docker compose up -d --build
Note: Ensure you port-forward to https://localhost:5000 to properly test the Product Hunt OAuth Redirect URI.
DMCPS MCP questions
What tools does DMCPS MCP provide?
It exposes 5 tools, including ph_graphql_query, read_file, write_file, list_directory, run_shell_command.
What tools are available for the AI to use?
The AI can perform GraphQL queries against Product Hunt, read and write files, list directories, and execute shell commands within the isolated workspace.