Dokploy MCP Server – Deployment Management Tools

Dokploy MCP Server exposes Dokploy API endpoints as Model Context Protocol tools for MCP-compatible clients. It is useful when an AI agent needs controlled access to manage Dokploy projects, applications, databases, deployments, Docker Compose resources, backups, domains, and related platform resources through a configured Dokploy instance.

#deployments#self-hosting#infrastructure

Overview

Dokploy MCP Server is the official MCP package for Dokploy, a self-hosted deployment and platform management system. The server connects MCP-compatible clients to a running Dokploy instance and exposes Dokploy API operations as tools. This lets an AI agent work with live deployment context instead of relying only on static documentation or copied logs.

What the MCP server enables

The MCP server exposes broad coverage of the Dokploy API, with tools generated from Dokploy's OpenAPI specification. The documented tool set covers project and application management, deployment history and queues, Docker Compose resources, domains, environments, databases, certificates, backups, registries, notifications, SSO, users, teams, and related server resources.

Typical agent workflows include listing projects, inspecting applications, checking deployment state, triggering lifecycle actions, reviewing database resources, or narrowing tool availability to selected categories with DOKPLOY_ENABLED_TAGS. Because the server can expose hundreds of tools, filtering is important when a client performs better with a smaller tool surface.

When to use it

Use this MCP server when an AI assistant should help operate a Dokploy-managed environment. Practical examples include asking an agent to find which project owns an application, inspect deployment history, review application domains, summarize configured databases, or assist with routine deployment operations. It is best suited for operators and developers who already run Dokploy and want MCP clients such as Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed, or similar tools to interact with it.

Connection and authentication

The default transport is local stdio using the published package @dokploy/mcp, typically started with npx -y @dokploy/mcp. The server requires DOKPLOY_URL and DOKPLOY_API_KEY so it can authenticate to the target Dokploy instance. Docker is also documented for local stdio operation.

Dokploy also supports an HTTP mode by running the server with --http or MCP_TRANSPORT=http. In that mode, the modern Streamable HTTP endpoint is /mcp, and legacy SSE endpoints are available at /sse and /messages for older clients. The HTTP server still needs upstream Dokploy credentials in the server environment.

Key considerations

Treat the Dokploy API key as a privileged operational secret. Store it in the MCP client's secret or environment configuration, not in prompts or shared files. The server includes optional custom headers for reverse-proxy scenarios and includes response-redaction settings such as DOKPLOY_REDACT_ENV and DOKPLOY_REDACT_FIELDS, which are useful when responses may contain environment variables, passwords, tokens, compose files, or other sensitive deployment data. Use the smallest practical set of tools and permissions for the agent's job.

Frequently Asked Questions

When should an AI agent use the Dokploy MCP Server?

Use it when an AI agent needs to work with a live Dokploy instance, such as inspecting projects, reviewing application configuration, checking deployment history, managing Docker Compose resources, or assisting with database and domain operations. It is intended for teams that already operate Dokploy and want MCP clients to interact with that environment.

What does the Dokploy MCP Server add to an AI agent's capabilities?

It gives the agent tool-based access to Dokploy API operations rather than only general deployment knowledge. The documented server exposes a large generated tool set covering projects, applications, deployments, domains, environments, databases, backups, registries, certificates, notifications, SSO, teams, users, and related Dokploy resources.

What can the AI agent access or manage through this server?

The agent can access and operate resources available through the configured Dokploy API key, including application lifecycle actions, deployment records, Docker Compose services, database resources such as PostgreSQL, MySQL, MariaDB, MongoDB, Redis, and other platform objects. Actual access depends on the Dokploy instance, API token privileges, and any tool filtering configured for the server.

How is authentication configured for the Dokploy MCP Server?

Authentication is configured with `DOKPLOY_URL` for the target Dokploy server and `DOKPLOY_API_KEY` for the Dokploy API token. Optional settings include `DOKPLOY_CUSTOM_HEADERS` for reverse proxies such as Cloudflare Access, plus timeout, retry, tool-filtering, and redaction environment variables. Secrets should be passed through environment variables or a client secret store, never hard-coded in shared configuration.

Which transport should be used for the Dokploy MCP Server?

Use `stdio` for desktop and command-line MCP clients that launch the server locally with `npx -y @dokploy/mcp`. Use Streamable HTTP when running the server as an HTTP service for web or remote client integrations; the documented modern endpoint is `/mcp`. Use the SSE endpoint only for older clients that still require the legacy MCP 2024-11-05 SSE flow.