The SportMind MCP server exposes SportMind's sports intelligence as tools for Claude Desktop and other MCP-compatible AI agents.
Confirmed working: 2026 UEFA Champions League Final — all five layers loaded, SMS 100, pre-match signal produced 48 hours before kickoff. Direction CORRECT. See sportmind.dev/first-record/
The server exposes SportMind's intelligence layers as callable tools. An AI agent (Claude Desktop, Claude Code, or any MCP-compatible client) calls the tools to load intelligence, produce pre-match signals, look up fan token context, and assess macro conditions.
The server does not call any API. It exposes tools. The model calls the tools. The tools load from the SportMind core library on your machine.
- Python 3.10 or higher
- Claude Desktop installed
- SportMind core library cloned
pip install mcp aiohttp
Step 1 — Clone both repositories into the same parent directory:
git clone https://github.com/SportMind/SportMind
git clone https://github.com/SportMind/sportmind-mcp-serverYour directory should look like:
parent/
├── SportMind/
└── sportmind-mcp-server/
Step 2 — Install dependencies:
cd sportmind-mcp-server
pip install -r requirements.txtStep 3 — Configure Claude Desktop:
Open Claude Desktop.
Go to Settings → Developer → Edit Config.
This opens claude_desktop_config.json.
Add the sportmind server block (see config/claude_desktop_example.json for the exact block with your paths).
Step 4 — Restart Claude Desktop.
Step 5 — Verify the connection:
In Claude Desktop type:
"Load SportMind and tell me the current library version and how many calibration records exist."
If the tools are connected you will see the hammer icon and SportMind will respond with the current state.
Produces a full pre-match signal with all five layers loaded.
Input: match details (teams, competition, venue, date, kickoff time) Output: direction, adjusted_score, sms, confidence_level, recommended_action, composite_modifier, modifiers_applied, flags (lineup_unconfirmed, macro_override_active, neutral_venue, path2_active)
Produces a signal for a single team or participant.
Input: team/participant, competition, context Output: team-level signal with athlete and form modifiers applied
Loads current macro intelligence context.
Input: optional — specific regime or regulatory jurisdiction query Output: active macro regime, modifier value, any override conditions, regulatory context
Looks up fan token context for a given club or ticker.
Input: club name or token ticker (e.g. "Arsenal" or "$AFC")
Output: token status, FTP model if active, supply mechanics, omnichain status, registry data
Produces a demand and sentiment snapshot for a fan token at a given moment.
Input: token ticker, event context Output: demand signal, occasion weight modifier, supply event risk assessment
2026 UEFA Champions League Final PSG vs Arsenal · Puskás Arena · Budapest
Pre-match signal produced T-48h:
| Field | Value |
|---|---|
| DIRECTION | PSG (Home) |
| ADJUSTED SCORE | 55 → 58 (T-2h) |
| CONFIDENCE | MEDIUM → MEDIUM-HIGH |
| ACTION | ENTER |
| SMS | 100 |
| $AFC PATH_2 | ACTIVE |
Result: PSG win 4-3 on penalties (1-1 AET) Direction: CORRECT ✓ Record #130 — sportmind.dev/first-record/
| Client | Status |
|---|---|
| Claude Desktop | ✓ Confirmed |
| Claude Code | ✓ Compatible |
| Any MCP-compatible client | ✓ Compatible |
Model: any — the server is model-agnostic. Tested with Claude Sonnet 4.6.
SMS measures how many SportMind layers are loaded and active.
| SMS | Meaning |
|---|---|
| 100 | All five layers loaded ✓ |
| 80 | Four layers — flag missing layer |
| < 80 | Do not act on signal without understanding which layers are absent |
Check SMS in every signal output. Do not act on a signal with SMS < 80 without understanding which layers are absent.
The MCP server is a thin tool layer over the SportMind core library. It does not contain intelligence — it exposes the intelligence in the core library as callable tools.
Keep both repositories updated. When the core library releases a new version pull the latest and restart Claude Desktop to ensure the tools reflect current intelligence.
MIT — same as the SportMind core library.
sportmind.dev · github.com/SportMind