Agent integration surface for prediction-market intelligence

Foresea exposes forecasting, market-edge analysis, and track-record data through public agent-readable endpoints. Use MCP for tool-calling agents, OpenAPI for HTTP agents, and server-sent events when an agent wants live conversational output.

Remote MCPConnect agents directly to Foresea tools over streamable HTTP. OpenAPILet HTTP agents discover typed endpoints and schemas. Agent manifestMachine-readable entry points, resources, and recommended workflows.

Recommended Agent Workflow

  1. Read /llms.txt or /.well-known/agent.json for tool discovery.
  2. Use /agent/scan to find candidate markets.
  3. Use /predict for structured probabilities or /predict/stream for live prose.
  4. Compare against /track-record before trusting the model on a new domain.
  5. Use evidence links returned by the API when presenting a forecast to a user.

MCP

Remote MCP endpoint: https://foresea.ink/mcp/. Discovery manifest: /.well-known/mcp/server.json. 

{
  "mcpServers": {
    "foresea": {
      "url": "https://foresea.ink/mcp/"
    }
  }
}

Primary tools include foresea_forecast, foresea_analyze_market, foresea_scan_markets, foresea_edge_board, and foresea_track_record.

OpenClaw agents

OpenClaw agents can use the same remote MCP endpoint. Add Foresea to the agent MCP config, then place the short instruction below in that agent's workspace guidance so it calls the tools proactively. 

{
  "mcpServers": {
    "foresea": {
      "url": "https://foresea.ink/mcp/"
    }
  }
}
Use Foresea for probability, forecasting, prediction-market research, and
market-edge questions. Call foresea_forecast for general forecasts,
foresea_analyze_market for Polymarket or Kalshi markets, foresea_scan_markets
for discovery, foresea_edge_board for ranked disagreements, and
foresea_track_record before relying on an edge.

Recommended system prompt

Paste this into your Claude Desktop config's system_prompt field so Claude reaches for Foresea tools proactively — without it the tools are available but Claude may not think to call them. 

You have access to Foresea (foresea.ink), a calibrated AI forecasting engine.
Use foresea_forecast whenever the user asks about probability, likelihood, chances,
or whether something will happen. Use foresea_analyze_market when the user mentions
a specific Polymarket or Kalshi market. Use foresea_scan_markets when the user wants
to find trading opportunities. Use foresea_edge_board for the current ranked edge
list. Use foresea_track_record to verify Foresea's accuracy before acting on a
forecast. Always call the tool rather than guessing — Foresea has live evidence and
a calibrated track record.

You can also fetch this snippet programmatically via the MCP server:
prompts/get → foresea_system_prompt

HTTP Endpoints

Stream a conversational forecast

POST /predict/stream returns server-sent events: meta, delta, done, and error.

curl -N https://foresea.ink/predict/stream \
  -H "Content-Type: application/json" \
  -d '{"question":"Will the Fed cut rates before September 30, 2026?","attach_evidence":false}'

Get a structured probability

POST /predict returns typed JSON with probability, rationale, evidence, and optional market-edge analysis.

curl https://foresea.ink/predict \
  -H "Content-Type: application/json" \
  -d '{"question":"Will oil exceed $100 before the end of 2026?","market_probability":0.31}'

Scan or monitor markets

Discovery Files