MCP Workflows #

Premium customers can use sfs-mcp with Claude Desktop, Claude Code, Cursor, or another MCP-capable assistant. These workflows show how a single natural-language request can trigger several Scottfree Sports tool calls.

The assistant supplies the reasoning and writing. Scottfree Sports supplies the data.

Tool Rules #

Public customer MCP tools:

get_daily_brief
get_predictions
get_results
get_summary
get_odds
get_sports
get_account_info
get_usage
get_kalshi_markets
get_polymarket_odds
compare_market_odds
get_game_weather
get_live_scores
get_injuries
get_players
get_betting_consensus

Sports:

mlb, nba, nfl, nhl, ncaaf, ncaab

Model types:

over_under, won_on_points, won_on_spread

Reading rules:

• Moneyline and spread probabilities are home-team relative.
• Totals probabilities are over-relative.
• Delta(Model - Implied) means model probability minus market-implied probability on that same axis.
• Line movement comes from opener/current fields in predictions and results.
• Scottfree Sports does not publish a guaranteed betting selector. Treat the tools as research data.

Workflow 1: Research Today’s MLB Totals #

Ask:

Research today's MLB totals. Use Scottfree Sports data as source data.
Show current total, opener-to-current movement, BLEND over probability,
implied over probability, Delta(Model - Implied), recent model summary,
weather if available, and the main uncertainty for each game.

Likely tool sequence:

1. get_daily_brief with {"sport":"mlb"}
2. get_predictions with {"sport":"mlb","model_type":"over_under"}
3. get_odds with {"sport":"mlb"}
4. get_summary with {"sport":"mlb","model_type":"over_under"}
5. get_game_weather with {"sport":"mlb"}

The assistant should return:

• matchup and ET game time,
• current total and opening total,
• BLEND over probability,
• implied over probability,
• signed model-vs-implied delta,
• recent over/under model context,
• weather context only when available,
• a clear note when the data is incomplete.

Workflow 2: Explain One NBA Spread Matchup #

Ask:

For tonight's Celtics/Lakers game, gather the NBA spread model prediction,
current odds, recent spread results, model summary, and injuries if available.
Explain what the model says and what I should verify before making a decision.

Likely tool sequence:

1. get_predictions with {"sport":"nba","model_type":"won_on_spread"}
2. get_odds with {"sport":"nba"}
3. get_results with {"sport":"nba","model_type":"won_on_spread","limit":50}
4. get_summary with {"sport":"nba","model_type":"won_on_spread"}
5. get_injuries with {"sport":"nba"} if a BALLDONTLIE GOAT-tier key is configured

The assistant should return:

• the game row it matched,
• home spread and away spread,
• opener/current movement,
• model probabilities,
• recent spread context,
• injury availability status,
• uncertainty and caveats.

Workflow 3: Compare Model Data With Markets #

Ask:

Find NHL moneyline games where Scottfree model probabilities disagree most
with available prediction-market prices. Show the model side, implied price,
Kalshi or Polymarket context if available, and whether the external market
is thin or missing.

Likely tool sequence:

1. get_predictions with {"sport":"nhl","model_type":"won_on_points"}
2. compare_market_odds with {"sport":"nhl","model_type":"won_on_points"}
3. get_kalshi_markets with {"sport":"nhl"}
4. get_polymarket_odds with {"sport":"nhl"}

The assistant should not pretend a missing or thin external market confirms the model. It should identify where the market data is usable and where it is not.

Workflow 4: Check Account And Quota #

Ask:

Check my Scottfree Sports account, plan, and remaining API quota.

Tool sequence:

1. get_account_info
2. get_usage

The assistant should return:

• active plan,
• subscription status,
• monthly request limit,
• requests used,
• requests remaining,
• reset date when available.

Workflow 5: Live Game Context #

Ask:

For today's NBA games, combine model predictions with live scores, player
availability, and public consensus. Keep each data source separate.

Likely tool sequence:

1. get_predictions with {"sport":"nba","model_type":"won_on_points"}
2. get_live_scores with {"sport":"nba"}
3. get_players with {"sport":"nba","per_page":25} if player lookup is needed
4. get_injuries with {"sport":"nba"} if injury access is available
5. get_betting_consensus with {"sport":"nba"}

The assistant should keep model data, live score data, injuries, and public betting data in separate sections so the customer can see which source supports each claim.

Workflow 6: Full Premium Slate Review #

Ask:

Run a full Premium research pass for today's MLB slate. Start with account and
quota, then use the daily brief, all three model markets, current odds, summaries,
recent results, public consensus, weather, and prediction-market context. Rank the
games by model-vs-implied disagreement and explain what each data source says.

Likely tool sequence:

1. get_account_info
2. get_usage
3. get_daily_brief with {"sport":"mlb"}
4. get_predictions with {"sport":"mlb","model_type":"over_under"}
5. get_predictions with {"sport":"mlb","model_type":"won_on_spread"}
6. get_predictions with {"sport":"mlb","model_type":"won_on_points"}
7. get_odds with {"sport":"mlb"}
8. get_summary for the relevant model types
9. get_results for the relevant model types
10. get_betting_consensus with {"sport":"mlb"}
11. get_game_weather with {"sport":"mlb"}
12. compare_market_odds with the most relevant model type

The assistant does not need to call every possible tool every time. A good answer uses enough tools to answer the question and says when a source is unavailable.

Good Prompts #

Show me today's MLB model picks ranked by Delta(Model - Implied). For each game,
include current odds, opener/current line movement, and recent model summary.

For NBA spreads, find games where the model and public consensus disagree.
Explain the model side, public side, current line, and recent spread performance.

Check my account quota, then summarize today's NHL moneyline model research with
available Kalshi and Polymarket context.

For NCAAF totals, show the over-relative model probabilities and explain which
games have the largest model-vs-implied deltas.

Bad Prompts #

Avoid prompts that ask the assistant to invent certainty:

Give me guaranteed winners.

Use Scottfree Sports to calculate a sure Kelly bet.

Better:

Show model disagreement, current prices, line movement, recent performance, and
uncertainties so I can make an informed decision.