Dreamhack MCP

Solving Dreamhack problems meant repeating the same manual steps each time: checking the list, downloading files, and deploying locally.

Registered login, fetching web problems by difficulty, downloading challenge files with automatic unzip, deploying via Docker or app.py, and stopping deployments as fastmcp tools.

The LLM handles fetching a problem and bringing it up locally through tool calls. I iterated through 49 commits in two days and deployed it to Smithery.