The repository states plainly that it is mid-restructure, and the honest read is that the champion, item and rune tools work today while much of the roadmap does not exist yet. What is implemented is genuinely hard to get elsewhere: stats at a chosen level, not just base values, which is what most build and damage questions actually need. Treat the analytics and live-match sections as intent rather than capability.
An MCP server that scrapes the League of Legends Wiki and serves the result as structured data. It covers champion statistics and abilities, item and rune details, and patch history for each of them, so a question about a champion's numbers is answered from the current wiki rather than a stale export.
- `get_champion_stats` returns a champion's statistics, with an optional level from 1 to 18 rather than base values only
- `get_champion_abilities` returns all abilities, and `get_ability_details` goes further into the mechanics behind one
- `search_champions` finds a champion with fuzzy matching, so a near-miss on spelling still resolves
- Items — `get_item_data` for statistics, recipe, cost analysis, notes and similar items
- Runes — `get_rune_data` for the sidebar, notes and strategy text
- Patch history for all three: `get_champion_patch_note`, `get_item_patch_note` and `get_rune_patch_note`, optionally scoped to one patch version
- `ping` and `server_info` for checking the connection and what the server is running
Python in a virtual environment, with the dependencies installed from the repository. The client launches the stdio server module — `python -m src.mcp_server.stdio_server` — from the project directory. Level-specific statistics are scraped with Selenium, so that path needs a browser driver available. Data comes from the wiki at request time, so the machine needs network access.