Agent-native sports and market data

Ask which market moved. Get an answer your app can use.

Connect one MCP endpoint. Your agent can find the relevant game, compare eligible prices, inspect history, and return stable IDs your application can reuse.

Connect an agent Choose your client
Tools
17
Access
Read-only
Identity
Stable IDs
Transport
Streamable HTTP
Agent trace preview
Replay · not live

Agent

Compare the same Arsenal outcome across public venues and retain timestamps and data-quality states.

get_market_event Read-only call

event_id=evt_ars_che_20260712

Data quality

Largest comparable gap

9.36 pp

Equivalence

Exact event · full game

Inspect the result path

The agent summarizes structured data. A human or application decides what to do with it.

01 · Find

Locate the game or signal

Search current games, qualified market gaps, recent form, or a specific event ID.

02 · Explain

Keep the useful context attached

Carry game state, outcome, timestamps, quality states, and history into the answer.

03 · Build

Continue in code with the same IDs

Move from the answer to REST, realtime, an alert, or your own customer-facing product.

Connect in minutes

One endpoint. Use the client you already work in.

The hosted MCP server is available now with a Scoreline project key. These are configuration snippets, not claims about a package or marketplace listing.

Set the project key once 
export SCORELINE_API_KEY='your_project_key'

Codex

Terminal command

Configuration available now

Register the hosted MCP server

Codex CLI, the Codex app, and the IDE extension share this MCP configuration on the same host.

Terminal command 
codex mcp add scoreline --url https://v2202608394083495819.ultrasrv.de/mcp --bearer-token-env-var SCORELINE_API_KEY

Claude Code

Project root · .mcp.json

Configuration available now

Add the project-scoped server config

Save this file in the project root. Claude Code reads SCORELINE_API_KEY from its environment, so the key is not copied into the file.

Project root · .mcp.json 
{
  "mcpServers": {
    "scoreline": {
      "type": "http",
      "url": "https://v2202608394083495819.ultrasrv.de/mcp",
      "headers": {
        "Authorization": "Bearer ${SCORELINE_API_KEY}"
      }
    }
  }
}

Cursor

~/.cursor/mcp.json

Configuration available now

Add the project or global config

Make SCORELINE_API_KEY available to the Cursor process, then enable the server in Cursor's MCP settings.

~/.cursor/mcp.json 
{
  "mcpServers": {
    "scoreline": {
      "url": "https://v2202608394083495819.ultrasrv.de/mcp",
      "headers": {
        "Authorization": "Bearer ${env:SCORELINE_API_KEY}"
      }
    }
  }
}

Windsurf

~/.codeium/windsurf/mcp_config.json

Configuration available now

Add the remote HTTP config

Use the raw MCP configuration, then refresh the MCP list so Cascade discovers Scoreline's tools.

~/.codeium/windsurf/mcp_config.json 
{
  "mcpServers": {
    "scoreline": {
      "headers": {
        "Authorization": "Bearer ${env:SCORELINE_API_KEY}"
      },
      "serverUrl": "https://v2202608394083495819.ultrasrv.de/mcp"
    }
  }
}

Any Streamable HTTP client

Connection contract

Protocol connection available now

Use these connection values

If your client accepts a remote MCP URL and bearer header, it can use the same project-scoped endpoint.

Connection contract 
transport=streamable-http
url=https://v2202608394083495819.ultrasrv.de/mcp
authorization=Bearer $SCORELINE_API_KEY

These clients connect directly to Scoreline's hosted MCP endpoint. No extra Scoreline package is required, and the project key remains in your environment.

MCP and Build with AI

The connection carries data. A recipe carries the workflow.

They reinforce one another, but they are not interchangeable and they do not create separate Scoreline products.

Available now

MCP is the live connection

MCP gives the agent authenticated, structured access to Scoreline tools at runtime. It carries current data, stable event IDs, quality states, project quota, and request traces.

  • Hosted Streamable HTTP endpoint
  • Callable read-only sports and Markets tools
  • Same project key and contract as the API
Recipes available now

Build with AI recipes carry the workflow

A recipe gives a coding agent the exact endpoints, Test event, expected result, security rules, and verification step for a repeatable Scoreline workflow.

  • Copyable instructions available today
  • Uses the same REST and MCP contracts
  • Optional installable packages can follow later

Focused, not fragmented

Three focused recipes. One shared data contract.

Copy a complete workflow into a coding agent today. Optional installable packages can follow without changing the API or MCP contract.

Copyable recipes available now
Copyable recipe

For coding agents building sports products

Live sports integration

A composed-first workflow for finding stable events, opening a consumer-ready outlook, and inspecting smaller domain inputs only when the product needs them.

  • Find stable sport, competition, and event IDs
  • Use one composed outlook or independently fetch the exact primitive needed
  • Return runnable requests and explicit missing-data states
Preview the operating prompt

Implement a Scoreline Test integration in this project. Contracts: - Operation: /docs/api/list-events - OpenAPI: /docs/openapi.json - Authentication: /docs/authentication Working Test request: GET /v1/events?limit=3 Header: x-api-key: $SCORELINE_API_KEY The response must include the stable event ID evt_ars_che_20260712. Requirements: - Read SCORELINE_API_KEY only on the server. Never hardcode, log, render, or send it to the browser or an AI prompt. - Implement loading, populated, empty, 401, 429, and retryable 5xx states. - Preserve stable event IDs when linking to event detail or another resource. - For a football match outlook, start with get_pre_event_outlook or GET /v1/analysis/events/{event_id}/pre-event-outlook. - If the UI also needs current score or state, load get_event independently once the event ID is known; do not create a sequential waterfall. - Use list_participant_events or get_market_consensus only when the UI needs those inputs separately. - Before get_market_consensus, call get_market_event and select the required market_id by matching market type, period, line, and settlement rules to the UI. - Use only capabilities advertised on the event response. Acceptance: - Run the real Test request; do not mock it. - Assert evt_ars_che_20260712 is rendered from the response. - Capture the response request ID and confirm the same request in Scoreline Request logs. - Report the files changed and the verification command you ran.

Copyable recipe

For research agents comparing public market history

Market history analysis

A bounded workflow for comparing compatible outcomes, querying historical windows, preserving data-quality states, and handing results back for human judgment.

  • Resolve event, market, outcome, and settlement equivalence
  • Compare aligned observations across named public venues
  • Describe price gaps without betting advice or execution
Preview the operating prompt

Build a bounded Scoreline market-history analysis using the Test environment. Contracts: - MCP setup: /docs/connect-an-agent - Pricing: /docs/api/get-event-market-pricing - History: /docs/api/get-market-history - OpenAPI: /docs/openapi.json Use SCORELINE_API_KEY only through the MCP client environment. Never place the key in source, output, or chat history. Exact Test calls: 1. compare_market_prices with event_id=evt_ars_che_20260712 and market_id=mkt_ars_che_full_game 2. Choose one returned outcome_id and the returned venue slugs, then call get_odds_history from=2026-07-12T20:00:00Z to=2026-07-12T23:45:00Z with granularity=1m Requirements: - Keep Scoreline's event, market, and outcome identities unchanged. - Show best displayed prices, venue hold, the named no-vig reference, timestamps, movement, data quality, and explicit missing intervals. - Treat theoretical arbitrage as descriptive math; fees, liquidity, limits, and execution are not assessed. - Do not recommend, rank, or execute a bet. Acceptance: - Confirm both responses retain the requested event and market IDs and history retains the selected outcome ID. - Capture the MCP request ID and confirm the same request in Scoreline Request logs. - Report assumptions, files changed, and the verification command.

Copyable recipe

For coding agents adding a monitored workflow

Price movement alert

A safe recipe for choosing one event and market condition, previewing the selection, and creating an alert through the supported application flow.

  • Choose an explicit event, outcome, and threshold
  • Preview current data before saving the alert
  • Keep consequential interpretation outside Scoreline
Preview the operating prompt

Implement a Scoreline price-gap alert flow against the Test environment. Contracts: - Preview operation: /docs/api/get-market-comparison - Create operation: /docs/api/create-market-alert - OpenAPI: /docs/openapi.json Test fixture: event_id=evt_ars_che_20260712 market_id=mkt_ars_che_full_game outcome_id=out_ars_che_home venues=kalshi,polymarket Requirements: - Keep SCORELINE_API_KEY and webhook secrets server-side and out of logs, browser code, and AI prompts. - Preview the current comparison before any write. - Let the user set a probability-point threshold and delivery destination. - Show loading, validation, unavailable-data, delivery, and retry states. - Require explicit confirmation immediately before POST /v1/markets/alerts. - Never create or imply a betting recommendation. Acceptance: - Verify the preview retains the exact event, market, and outcome IDs. - Verify cancel performs no write and confirm performs one authenticated write. - Capture the response request ID and confirm the same request in Scoreline Request logs. - Report the files changed and the verification command.

Callable now

A compact tool catalog for real product work.

Tools stay factual and composable. The agent retrieves exact data and quality context; it does not narrate a betting recommendation or execute an action.

Sports product tools

9 tools

Discover, inspect, and integrate stable sports data

  • search_sports_data Discover stable sports entities.
  • list_live_events List live events with compact scores and state.
  • list_schedule Find scheduled events in a bounded range.
  • get_event Read one event, freshness, and capabilities.
  • get_event_resource Read an advertised event resource.
  • list_participant_events Read finished-event history before a time boundary for one participant.
  • get_standings Read competition standings.
  • get_coverage Check current capability and history support.
  • search_docs Find the relevant API operation or guide.

Analysis tools

1 tool

Open a consumer-ready outlook without assembling its inputs yourself

  • get_pre_event_outlook Read the composed football outlook with quality and evidence.

Market research tools

7 tools

Inspect compatible public prices, history, and data quality

  • get_market_event Inspect one event with comparable public markets.
  • get_market_consensus Read no-vig probabilities and explicit quality status for one market.
  • get_odds_history Retrieve bounded probability history.
  • compare_market_prices Compare current equivalent probabilities.
  • get_closing_line Read the last eligible pre-event quote.
  • find_market_divergence Find the largest compatible price gaps.
  • resolve_market_contract Inspect event identity and market compatibility.

Deterministic replay · not live

Verify the connection with a result that can be reproduced

The public replay keeps the event, comparison, observation times, and caveats stable. It verifies the integration path without pretending to show a live opportunity.

  • The response retains the stable event ID
  • Comparable prices remain attached to the selected outcome
  • Timestamps and collection caveats remain inspectable
  • The agent leaves interpretation to a human or application
Open the human-readable replay
Verification prompt

Use Scoreline's get_market_event tool for evt_ars_che_20260712. Summarize the event, report current compatible prices for the selected outcome, include observation times and any collection caveat, and treat price gaps as data rather than recommendations.

Equivalent MCP call 
{
  "name": "get_market_event",
  "arguments": {
    "event_id": "evt_ars_che_20260712",
    "venues": "bookmaker_consensus,kalshi,polymarket"
  }
}

The responsibility chain

Agent to structured data to human.

Scoreline makes retrieval and verification easier without moving judgment into a hidden model response.

  1. 01

    Agent requests a bounded fact

    It finds a stable object, then calls an exact tool with explicit IDs, venues, and time boundaries.

  2. 02

    Scoreline returns structured data

    Stable IDs, observation times, freshness, completeness, and relevant caveats travel with the result.

  3. 03

    A human or application decides

    The agent can summarize and hand off stable IDs, but judgment and consequential action stay outside Scoreline.

Secure by product boundary

Useful access without an execution surface.

MCP reuses the same project identity and observability as the API. The public tool catalog remains read-only and data-focused.

Read-only tools

The public MCP catalog retrieves sports data and quality details; it does not mutate Scoreline state.

Project-scoped access

Every call uses a project key, its environment, quota, and allowed read scope.

Shared request traces

MCP activity enters the same request-log and usage path as the API.

Uncertainty stays explicit

Missing capabilities, stale observations, and caveats are returned instead of inferred.

No picks or execution

Tools report sports and market data; they do not recommend bets or place orders.

One genuine tool call is the activation

Connect the client. Run the replay. Inspect the trace.

Create a project key, add the hosted MCP endpoint, and continue in the agent workspace with the same Test event, configuration, and request logs.

Start with a project key Read MCP documentation