Skip to main content

mittwald MCP

Prompt your agent with intent. Let mittwald MCP execute the infrastructure work with safe, auditable tool calls.

An example of using mittwald MCP with an AI assistant to manage infrastructure.

Run first workflow

Connect your AI assistant to mittwald MCP and run your first workflow.Learn more
‍

Documentation

Learn about authentication, the available tools and step-by-step tutorials.Learn more
‍

GitHub

Browse the source code of the MCP server and report issues.Learn more
‍

Just want to get started quickly? Use this MCP URL in the AI assistant of your choice:

https://mcp.mittwald.de/mcp

See our setup guides‍ for more detailed instructions.

Agent-first infrastructure

mittwald MCP connects your AI coding assistant to mittwald's infrastructure. Describe what you want in natural language, and let your agent handle the API calls.

Comprehensive toolset

Full coverage across projects, apps, databases, DNS, mail, security, and automation. One MCP endpoint for all mittwald operations.

  • Project and app management
  • Database provisioning
  • Domain and SSL configuration
  • Backup and restore

Flexible authentication

Choose the auth method that fits your workflow:

  • OAuth 2.1 for interactive agents with automatic token refresh
  • API tokens for CI/CD pipelines and headless environments

Setup guides →‍

Get started

Connect your preferred AI assistant to mittwald MCP and start managing infrastructure with natural language prompts.

Connect your assistant

Step-by-step setup guides for popular AI coding tools:

Learn the pattern

Human intent in prompts, MCP tools for execution, explicit approvals for risky actions.

How it works

Agent-driven workflows follow a consistent pattern: you provide intent, the agent discovers and executes tools, you verify results.

The agent workflow

  1. Intent — You prompt the outcome you want on mittwald
  2. Discovery — Your agent discovers relevant MCP tools and required inputs
  3. Auth — OAuth or token auth is used for each tool call
  4. Execution — The agent performs tool calls and reports results
  5. Verify — You run read checks and decide whether to proceed, retry, or roll back

Human responsibility: Destructive or cost-impacting changes should be approved explicitly. Treat the agent as an operator, not as an unattended process.

Tutorials by team

Outcome-driven walkthroughs organized by team context and use case.Browse all tutorials →‍

Freelancers & agencies

  • Client onboarding automation
  • Cross-project visibility
  • Team access management
  • Backup monitoring

E-commerce & DevOps

  • Launch day preparation
  • Database performance checks
  • Container stack deployment
  • CI/CD pipeline integration