RiotSkillIssue MCP server

RiotSkillIssue

Typed async and sync Riot Games APIs, plus an optional local MCP server.

Documentation · Migration guide · API reference

Install

RiotSkillIssue 1.x requires Python >=3.14,<3.17. CI tests Python 3.14 and 3.15 on Linux and Windows, and release verification checks both versions. A separate Python 3.16-dev compatibility probe runs when its dependencies can be installed; unavailable upstream dependencies are reported without failing that probe.

pip install riotskillissue

Set RIOT_API_KEY or pass a key directly:

import asyncio

from riotskillissue import PlatformRoute, RiotClient


async def main() -> None:
    async with RiotClient(
        api_key="RGAPI-...",
        default_route=PlatformRoute.EUW1,
    ) as riot:
        profile = await riot.lol.player_profile("Player#EUW")
        history = await riot.lol.match_history("Player#EUW", count=5)
        match = await riot.raw.lol.match.get_match(match_id="EUW1_...")
        print(profile, history, match)


asyncio.run(main())

The explicit synchronous client provides the same workflows and raw operation groups:

from riotskillissue import PlatformRoute, SyncRiotClient

with SyncRiotClient(default_route=PlatformRoute.EUW1) as riot:
    print(riot.lol.player_profile("Player#EUW"))

Replay downloads

Open the League Client and sign in, then choose a replay from another player's recent matches using their full Riot ID:

riotskillissue-cli replays list --player "GameName#TagLine"
riotskillissue-cli replays download --player "GameName#TagLine"

For a direct selection and an explicit destination:

riotskillissue-cli replays download 1 --player "GameName#TagLine" --output ./replays/my-game.rofl

This requires League running on the same Windows or macOS machine, with the target match accessible on the client's region. No API key or RSO token is needed. Riot's limit of 10 downloads per hour applies through the client. The League Client API is not officially supported by Riot and may change between patches.

The public API workflow remains available by omitting --player and setting RIOT_RSO_ACCESS_TOKEN. It returns replays for that token's player. Python applications can use LeagueReplayClient for player lookups or the async/sync lol.replays() and lol.download_replay() RSO helpers. See the replay guide for client setup, RSO authentication, limits, and examples.

Local MCP server

Install the optional server and run it over stdio:

pip install "riotskillissue[mcp]"
riotskillissue-mcp

The server reads credentials from its environment. API keys and RSO tokens are never tool arguments. Raw write operations are hidden unless RIOT_MCP_ALLOW_WRITES=true; enabled writes still require human confirmation.

League-aware tools assemble Match V5, timeline, player, ranked, mastery, and patch-matched Data Dragon evidence for natural-language questions. For example, riot_lol_match_context can resolve a match directly or from a player's recent history and assess Void Grub captures, trades, and later turret conversion while clearly separating observed telemetry from causal inference. Patch-banded fundamentals cover minion-wave gold and XP, sharing, passive income, structures, objectives, role modifiers, and wave-management opportunity costs. riot_lol_item_economy derives patch-matched raw-stat efficiency from Data Dragon component prices while leaving passives and unrepresented stats unpriced. See the MCP guide for tools, production limits, and data-coverage boundaries.

Coverage and generation

The generated raw API covers every operation in the bundled community-maintained OpenAPI feed. That feed is not an official Riot Games specification. One committed operation registry drives raw dispatch, reference documentation, and MCP discovery.

RiotSkillIssue is not endorsed by Riot Games and does not reflect the views or opinions of Riot Games or anyone officially involved in producing or managing Riot Games properties. Riot Games and all associated properties are trademarks or registered trademarks of Riot Games, Inc.

License

MIT. See LICENSE.