rivalsdata-api

An unofficial Python client for RivalsData's public Marvel Rivals data. It uses the site's undocumented API, so routes and fields can change. The client keeps unknown response fields accessible instead of discarding them.

Install

Python 3.10 or newer:

python -m pip install rivalsdata-api

For editable development, clone the repository and run python -m pip install -e '.[dev]'. The optional Camoufox Cloudflare fallback is installed with python -m pip install 'rivalsdata-api[browser]', followed by python -m camoufox fetch.

Quick start

from rivalsdata import RivalsDataClient

with RivalsDataClient() as rd:
    player = rd.get_player("GS-")  # numeric UID works too
    print(player.name, player.level, player.rank_game_season)

    # Player profile sections are lazy resource managers.
    hero_season = player.heroes.fetch(season=20)
    map_stats = player.stats.maps(season=20)
    match_page = player.matches.fetch(season=20)

    # Current match (None if the profile is not currently in a game).
    live_game = player.live_game.fetch()
    if live_game is not None:
        print(live_game.players, live_game.team_avg_rank)

    # Site-wide resources are available from the client.
    leaderboard = rd.leaderboards.fetch(limit=100, season=20, platform=1)
    tier_list = rd.heroes.tier_list(platform=1, rank="grandmaster_plus")
    team_ups = rd.team_ups.fetch(platform=1, rank="grandmaster_plus")
    xp_page = rd.insights.xp()

print(hero_season[0].win_rate)  # integer percent when wins/losses are present

MCP server (ChatGPT and Claude)

Install the MCP extra and the package:

python -m pip install 'rivalsdata-api[mcp]'

The server exposes read-only tools for player search and profiles, a player's current live match (when they are in one), match history, player stats, leaderboards, heroes, team-ups, public insights, matches, and factions. The show_player_dashboard tool also returns an MCP-UI player report with rank and competitive record, current match roster split by side, a hero win-rate chart, and recent match form with K/D/A. It supports local stdio for Claude Desktop and Streamable HTTP for remote MCP clients such as ChatGPT. Data comes from RivalsData's undocumented API and may change; profile match history can be private.

How the MCP UI works

show_player_dashboard fetches current data, then returns an HTML UI resource alongside the tool result. MCP-UI labels it with a ui:// resource URI and preferred size. A compatible host can render that resource in a sandboxed panel; a host without UI support can still use the regular MCP tools and their text/data responses. ChatGPT uses MCP-UI's Apps SDK adapter, while Claude is listed as supporting MCP Apps directly. The dashboard is a snapshot from the time the tool runs; ask for it again to refresh.

Claude Desktop (local)

Add a server entry to Claude Desktop's claude_desktop_config.json, replacing the path with the Python executable in the environment where the extra is installed:

{
  "mcpServers": {
    "rivalsdata": {
      "command": "C:\\path\\to\\venv\\Scripts\\python.exe",
      "args": ["-m", "rivalsdata.mcp_server"]
    }
  }
}

On macOS/Linux, use the environment's bin/python path. Restart Claude Desktop after saving the configuration.

ChatGPT or remote Claude connector

Run the server on a host reachable over HTTPS:

rivalsdata-mcp --transport streamable-http --host 0.0.0.0 --port 8000

The MCP endpoint is /mcp (for example, https://your-host.example/mcp). Add that endpoint through the client's custom/remote MCP connector settings. The server does not implement authentication; put it behind an authenticated HTTPS gateway before exposing it publicly. For local development, bind to 127.0.0.1 instead. Use python -m rivalsdata.mcp_server --help to see options.

Player and returned DataModel objects support both mapping access and attribute access (player["level"] or player.level). Nested dictionaries and arrays are wrapped recursively; .raw returns a shallow copy of a model's original JSON. For endpoints whose fields evolve, these generic typed wrappers preserve the complete payload.

Public resources

• rd.leaderboards.fetch(...) — global player ranking.
• rd.heroes.tier_list(...), .get(hero_id), .meta(hero_id, range=90), .leaderboard(hero_id, **filters) — hero metrics and ranking.
• rd.team_ups.fetch(...) — team-up stats.
• rd.insights.punishments(...), .xp(...), .top_500(...), .commbans(...), .leavers(...) — public insights and cursor metadata.
• rd.factions.get(faction_id), rd.matches.get(match_id), rd.profiles.get(username), and rd.favorites.fetch(uids) — detail/profile lookups.
• player.heroes.fetch(...), .matches.fetch(...), .live_game.fetch(), .teammates.fetch(...), .crosshairs.fetch(), .proficiency.fetch(), .punishments.fetch(), .name_history.fetch() — profile sections.
• player.stats.heroes(...), .maps(...), .bans(...) — detailed profile stats.

See the observed API inventory for methods, parameters, observed response shapes, and endpoints that require a RivalsData account. The API inventory distinguishes observed behavior from inferred/unverified details.

Cloudflare fallback

Requests use curl_cffi with a Chrome TLS profile by default. If blocked, enable the optional browser fallback:

with RivalsDataClient(use_browser_fallback=True) as rd:
    player = rd.get_player(1970288503)

Errors and contributions

All package exceptions inherit from RivalsDataError. See CONTRIBUTING.md for setup, code layout, change workflow, and notes for new contributors. docs/PROJECT_CONTEXT.md is the handoff document for new coding sessions.

This project is not affiliated with RivalsData, NetEase, or Marvel. Keep request rates reasonable and respect the site's terms.