FPL Strategy MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| FPL_MCP_BEARER_TOKEN | No | A bearer token used to protect the HTTP/streamable transport when running in remote mode (e.g., `fpl-strategy-mcp --transport streamable-http`). If set, the health endpoint and MCP endpoint require this token. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| fpl_recommend_movesA | Return legal FPL transfer moves for the current gameweek using the frozen champion strategy by default. Give a full 15-player current squad. buyable_players is optional: when omitted, the server loads the full current official player pool from the cached bootstrap snapshot/API. Give point-in-time signals or enable auto_official_signals, bank/free transfers, unused chips, and optional mini-league standings context. The default champion ranks hold and legal free transfers; explicit hybrid challengers can rank Wildcard, Free Hit, Bench Boost, and Triple Captain when model-backed. Use weight_overrides to tune the transparent decision layer per request. |
| fpl_lineup_planA | Choose the legal current-week formation, starting XI, bench order, captain and vice-captain from the supplied 15-player squad. It also reports the projected bench points that Bench Boost would add; normally only the XI scores. |
| fpl_search_playersA | Search the full cached official FPL player pool by name, position, team, price, and availability. Use this to construct or inspect buyable candidates; it returns current price, official expected points, form, ownership, status, and transfer signals. |
| fpl_forecast_signalsA | Return the point-in-time signal table for the supplied squad and buyable pool, including short/long expected points, future price signals, minutes, risk, news/social context, ownership leverage, and uncertainty. Use it to inspect the underlying inputs before asking for a decision. |
| fpl_score_movesA | Rank legal one-transfer moves under explicit transparent weights. This is the tuning and explanation tool for short/long points, price economics, ownership leverage, availability, news risk, uncertainty, hit cost, and risk aversion; use fpl_recommend_moves for the final action including chips and the learned policy. |
| fpl_strategy_infoA | Return the selected strategy, benchmark status, research sources, and known limitations. |
| fpl_strategy_catalogA | Return all available strategy profiles, action kinds, default decision/cocktail configs, signal components, and the fields that can be tuned per request or in a backtest. |
| fpl_backtest_strategyA | Backtest and compare tunable strategies. Use scenarios/scenarios_path for point-in-time replay with realized action_outcomes, or history_root plus season for the legal Vaastav-format season simulator. To reproduce the published 2,486-point research candidate, use forecast_model=external_hgb, policy=patient_chips, and initial_squad_modes=[forecast]. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 8 tools
Most tools have clearly distinct responsibilities: search, forecast, lineup, backtest, catalog, and info are well-separated. The main ambiguity is between fpl_recommend_moves and fpl_score_moves, though the descriptions do clarify that one is for final policy-driven actions and the other for transparent weight tuning of one-transfer moves.
The consistent fpl_ prefix and snake_case help readability, and most names follow a verb_noun pattern like fpl_recommend_moves, fpl_search_players, fpl_score_moves. However, fpl_lineup_plan, fpl_strategy_catalog, and fpl_strategy_info deviate from that pattern, mixing noun-led and verb-led naming.
Eight tools is a well-scoped count for an FPL strategy server, covering search, signals, decision-making, lineup planning, and backtesting without bloat. Each tool appears to serve a distinct part of the analysis-to-action workflow.
The server covers the main FPL workflow: player discovery, signal inspection, move scoring, final recommendations, lineup selection, and strategy backtesting. Minor gaps exist, such as no explicit tool for creating or persisting custom strategies or for managing chip state directly, but these are largely addressable through existing tools.