Gatekeeperd MCP is an MCP server built around Gatekeeperd, a VPS deployment and infrastructure management platform. It exposes infrastructure operations as structured tools that AI agents can discover and invoke instead of relying on SSH sessions or manually navigating a deployment dashboard.
The MCP server currently provides tools across multiple areas of Gatekeeperd, including project management, deployments, Docker containers, Docker Hub, domains, Nginx, runtime health, credentials, billing, and infrastructure operations. This allows an AI agent to move beyond generating deployment instructions and actually interact with the infrastructure through controlled backend operations.
The architecture keeps the AI agent separated from the underlying infrastructure. The agent communicates with the MCP layer, while Gatekeeperd remains responsible for authentication, authorization, validation, business logic, deployment orchestration, and communication with external services.
A typical workflow can start with an agent inspecting a project, checking its deployment state, triggering a deployment, monitoring the resulting container, and inspecting runtime or reverse-proxy status. Deployment operations can also be designed around health checks so that a new container is validated before replacing a currently running instance.
The project is built with Kotlin and Ktor and integrates with Docker, Docker Hub, Nginx, Linux VPS infrastructure, GitHub, PostgreSQL, and Paystack. The MCP interface supports HTTP/SSE as well as local stdio usage, making the same infrastructure capabilities accessible to different types of AI clients and agents.
The larger goal is to make infrastructure programmable for both humans and AI agents. Instead of treating deployment as a sequence of shell commands or UI interactions, Gatekeeperd MCP exposes infrastructure as a set of explicit, composable capabilities that an agent can reason about and operate through.
Top comments (0)