The useful part is not the search — it is that a guide comes back as Markdown from the article body, without the ad boxes, so a model reads the steps rather than the page furniture. Scope is exactly one site, aimed at beginners and children, and the URLs are checked before they are returned, so a suggestion that turns out to be a dead link is unlikely. The content belongs to that site: this is a reading path for personal and educational use, not a bulk copy.
A single-file Python MCP server over Easy Drawing Guides. It searches the site and validates every URL before returning it, fetches a chosen tutorial and converts the article body to Markdown with the ad blocks removed, and lists the site's categories for when you do not have a subject yet.
- A ranked list of matching tutorials with titles and URLs that have been checked for reachability before being handed back, with a configurable result limit — `search`
- One tutorial's full content as Markdown, taken from the article section with the ad boxes stripped, plus title, source and content length — `get_guide`
- The site's categories and popular search terms, as a starting point for browsing — `list_categories`
Python 3.12 or higher and the uv package manager. The README's `uvx` line names `mcp-drawing-guides`, which PyPI does not have; from a checkout, `uvx --from . mcp-drawing-guides` or `uv sync` then `uv run python main.py`. stdio only, no key and no account. It fetches pages live, so it needs network access, and its parsing follows the site's current article structure. The project version is 0.1.0.