This server lets Claude manage a Liga Record fantasy football squad: validate lineups, simulate automatic substitutions, plan transfers, search the player market, project prices, and check fixtures.
Squad management: get_squad, get_player, search_squad, squad_value, list_coaches, club_strength — read your 23-player squad, filter by club/position/value/points, and inspect coaches.
Rules enforcement: validate_selection checks formations, bench order, captain/coach rules against the regulation; simulate_autosubs applies automatic substitution rules.
Transfers: check_transfer and check_market_transfer evaluate swaps using budget and regulation rules; live market data prices incoming players.
Market research: search_market scans all players with ownership percentages to find differentials; search_squad filters only owned players.
Fixtures: get_fixtures and squad_fixtures provide the league calendar and each player's next opponent.
Projections: project_price estimates quote changes; project_points blends form with priors (not validated).
Player history: record_appearances and appearance_history give appearance data.
Resources & prompts: access the regulation and squad as resources, plus pick_starting_xi and plan_transfers prompt templates.
Live data awareness: reads carry as_of timestamps, and market results are cached; the server is read-only by design — buy/sell actions remain manual on the site.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@liga-record-mcp Is my current XI legal, and who comes on if Diogo Costa doesn't play?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP server I built to learn how the Model Context Protocol works, on a problem I actually have: managing my team in Liga Record, the fantasy football game of Record, a Portuguese sports newspaper. The design rests on one split: the rulebook is arithmetic, so it lives in Python as pure functions (legal formations, the €40M budget, automatic substitutions), while judgement, like who to start or who to sell, stays with Claude. Claude gets 27 tools, 3 prompts for the weekly routine and the regulation as a resource generated from the same constants the code enforces, and every read carries an as_of timestamp so Claude can tell stored data from live data. The server is read-only on purpose: the site's buy and sell endpoints are known but not implemented, and a test checks it stays that way, so a transfer is always a human click. It later grew a points model, and a scheduled GitHub Action records its predictions before every round and scores them afterwards, so the model is only judged on calls made in advance.
python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -e ".[dev]"
.\.venv\Scripts\python.exe -m pytest -qThe editable install creates the liga-record-mcp launcher that .mcp.json points at, so Claude Code starts the server when you open this folder. That path is absolute, so edit it after cloning. The squad lives in data/squad.yaml (start from data/squad.example.yaml), and the loader checks it against the regulation on every read. Then ask something like "Is my current XI legal, and who comes on if Diogo Costa doesn't play?"
More on the design in docs/HOW-IT-WORKS.md.
Check a real transfer: one squad player out, one market player in.
Both are player ids. Unlike check_transfer, the incoming player's price and position come from the live market rather than being described by hand — so the budget arithmetic uses their actual quote.
Inputs: in_id (required), out_id (required). Returns: object
Check one swap against §6.4 and §6.8.
The incoming player is described by hand because there is no market data
yet — step 4 replaces these arguments with a lookup. window is one of
in_season, closed (February) or reopened.
Inputs: out_id (required), in_club (required), in_name (required), in_value (required), +3 more. Returns: object
Find one squad player by id, or by name (partial and accent-insensitive).
Returns the match, or every candidate when the query is ambiguous — never a guess, since acting on the wrong player is worse than asking again.
Inputs: query (required). Returns: object
The 23 players under contract, with the money around them.
Carries as_of: say how fresh the data is rather than presenting a stored
squad as live.
Inputs: no named parameters. Returns: object
What a player's quote does if they score round_points (§12.3-§12.4).
Scores of 1, 2 or 3 are not covered by the regulation and are treated as no movement — say so if it matters to the answer.
Inputs: player_id (required), round_points (required). Returns: object
Search the whole Liga Record player market, not just the squad.
position is required — one of GK, DEF, MID, FWD. The site's endpoint
returns nothing without one.
max_owned_percent finds differentials: players few other teams hold. A
high-scoring player owned by 40% of the league gains you nothing on the
field; the same player owned by 3% is where places are won.
This one reads the live site, so it is the only tool here whose answer can change without anyone editing a file. Results are cached for 15 minutes — quotes only move when a round is scored.
Inputs: position (required), club, name, limit, +3 more. Returns: object
Filter the squad. position is one of GK, DEF, MID, FWD.
This searches the 23 players already owned, not the transfer market — the market needs the live source that step 4 adds.
Inputs: club, position, max_value, min_points. Returns: object
Apply the §11 automatic substitutions to a team sheet.
unavailable is the starters who did not play, or whose match was
abandoned or postponed after the round closed. Substitutes come on in bench
order, same position only, three at most, and a replaced captain passes the
armband on.
Inputs: bench (required), starters (required), unavailable (required), captain. Returns: object
Check a team sheet against §6.13, §6.17 and §10.3(l).
Bench order matters — the first substitute listed is the first to come on. This is authoritative: trust it over your own count of the formation.
Inputs: bench (required), starters (required), captain, coach_id. Returns: object
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 9 tools
Each tool has a clear and distinct purpose. check_transfer vs check_market_transfer and search_squad vs search_market are differentiated by their descriptions, preventing confusion. The others (get, validate, simulate, project) are uniquely scoped.
All tool names follow a consistent verb_noun pattern (check_transfer, get_squad, search_market, etc.), using lowercase with underscores. No mixed conventions or inconsistent verb usage.
Nine tools cover the core operations of a fantasy football assistant without redundancy. Each tool serves a distinct function, and the count is neither too sparse nor overwhelming for the domain.
The toolset covers reads, searches, validation, simulation, and pricing projections, but lacks an explicit 'execute transfer' or 'update squad' action. However, given the assistant's advisory role, this is a minor gap that doesn't hinder typical workflows.
Hosted MCP server for live public-data APIs and Skills for AI agents.
MCP server providing access to the Scorecard API to evaluate and optimize LLM systems.
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
MCP server giving Claude AI access to 22+ NYC public-record databases for real estate due diligence
