World Monitor MCP Server已验证已精选

Live, source-attributed global intelligence for AI assistants and autonomous agents.

World Monitor exposes its global-intelligence platform through the Model Context Protocol (MCP). Connect Claude, Cursor, MCP Inspector, or another MCP-compatible client to one hosted endpoint and give your agent structured access to markets, geopolitical risk, conflicts, aviation, maritime activity, energy, climate, cyber threats, supply chains, public health, sanctions, forecasting, and regional intelligence briefs.

Remote endpoint: https://worldmonitor.app/mcp
Transport: Streamable HTTP
Protocol: MCP 2025-06-18, with compatibility for 2025-03-26 clients
Server version: 1.16.0
Tools: 63 read-only tools
MCP Apps: 10 interactive views
Authentication: OAuth 2.1 or a World Monitor API key

Why World Monitor?

Most agents can search the web, but search results alone do not provide a stable operating picture. World Monitor combines hundreds of official, institutional, commercial, and open-source providers into normalized intelligence surfaces that agents can query consistently.

The MCP server is designed for analysis rather than unrestricted browsing:

• One global-intelligence endpoint across geopolitical, economic, security, movement, infrastructure, environmental, and humanitarian domains.
• Source-attributed results that preserve the underlying publishers, timestamps, corroboration details, and provenance signals available in each dataset.
• Explicit freshness metadata through cached_at and stale fields, so an agent can distinguish a current observation from an old or degraded snapshot.
• Structured output schemas for every tool, allowing capable clients to reason about response shapes before making a call.
• Server-side JMESPath projection on every tool to reduce large responses before they enter the model context.
• Read-only tool annotations with no destructive or state-changing agent operations.
• Public discovery for tools, prompts, and resources before authentication.
• OAuth 2.1 with PKCE for supported interactive clients, plus direct API-key authentication for scripts and custom integrations.

Intelligence coverage

Domain	Representative capabilities
Geopolitics and security	Conflict events, country risk, military posture, sanctions, unrest, cyber threats, defense-industrial dependencies, and escalation hotspots
Markets and economics	Equity and sector performance, commodities, crypto, foreign exchange, macroeconomic indicators, tariffs, trade flows, government debt, housing, industrial production, and consumer prices
News and OSINT	AI-classified news intelligence, corroborated story clusters, keyword spikes, entity extraction, social velocity, alert digests, timelines, and historical-event similarity
Aviation and maritime	Live airspace activity, military aircraft, airport disruption, flight search, vessel activity, maritime chokepoints, congestion, and route-risk intelligence
Energy and supply chains	Energy supply and storage, disruption monitoring, commodity geography, mineral production, procurement opportunities, trade flows, and infrastructure-cascade simulation
Climate and disasters	Earthquakes, wildfires, extreme-weather alerts, climate anomalies, radiation readings, environmental indicators, and population exposure
Health and humanitarian	Disease outbreaks, air quality, displacement, refugee and internally displaced population data, food stocks, and humanitarian risk signals
AI intelligence	World and country briefs, situation analysis, signal convergence, focal-point detection, forecasts, forecast scorecards, event classification, and entity extraction

Tool catalog

World Monitor currently exposes 63 tools. Use tools/list for the authoritative live schemas and describe_tool when you need the complete uncompressed definition of one tool.

Global risk, conflict, and security

• get_world_brief
• get_country_brief
• get_country_risk
• get_conflict_events
• get_military_posture
• get_military_surge
• get_hotspot_escalation
• get_sanctions_data
• get_cyber_threats
• get_defense_industrial_base
• get_china_decision_signals
• get_test_site_seismicity

News, alerts, and analytical signals

• get_news_intelligence
• get_news_clusters
• get_keyword_spikes
• get_social_velocity
• get_temporal_anomalies
• get_signal_convergence
• get_focal_points
• get_alert_digest
• search_intel_history
• get_intel_timeline
• get_similar_events
• classify_event
• extract_entities

Markets, economics, trade, and companies

• get_market_data
• get_economic_data
• get_country_macro
• get_eu_housing_cycle
• get_eu_quarterly_gov_debt
• get_eu_industrial_production
• get_prediction_markets
• get_tariff_trends
• get_wto_trade_flows
• get_procurement_opportunities
• get_food_stocks
• get_consumer_prices
• get_company_intelligence
• get_commodity_geo
• get_mineral_production

Aviation, maritime, and infrastructure

• get_aviation_status
• get_airspace
• get_maritime_activity
• get_chokepoint_status
• get_infrastructure_status
• get_supply_chain_data
• simulate_infrastructure_cascade
• search_flights
• search_flight_prices_by_date

Environment, energy, health, and humanitarian data

• get_energy_intelligence
• get_climate_data
• get_natural_disasters
• get_radiation_data
• get_health_signals
• get_displacement_data
• get_population_exposure
• get_positive_events
• get_research_signals

Forecasting and synthesis

• get_forecast_predictions
• get_forecast_scorecard
• analyze_situation
• generate_forecasts
• describe_tool

Built-in workflow prompts

The server includes six reusable MCP prompts for common intelligence workflows:

• country-briefing — assemble a structured country-risk and macro briefing.
• energy-shock-watch — evaluate energy disruption, supply, price, and route signals.
• market-open-prep — prepare a cross-asset and sector briefing before a market session.
• conflict-pulse — summarize conflict, military, and news escalation signals.
• route-risk-check — assess chokepoint, maritime, aviation, and supply-chain risk.
• freshness-audit — inspect the recency and degraded state of contributing datasets.

Interactive MCP Apps

Clients that support the io.modelcontextprotocol/ui extension can render ten interactive views directly from World Monitor tools:

• Country risk
• World brief
• Country brief
• Market radar
• Maritime chokepoint monitor
• News intelligence
• Conflict events
• Natural disasters
• Prediction markets
• Forecasts

The UI shells are static, public, and quota-exempt. Live data reaches them only through the normal authenticated tool call selected by the user or agent.

Connect with OAuth

Modern MCP clients can connect directly to the hosted endpoint:

{
  "mcpServers": {
    "worldmonitor": {
      "url": "https://worldmonitor.app/mcp"
    }
  }
}

On the first data-bearing call, a compatible client can start the World Monitor OAuth 2.1 flow. OAuth uses PKCE and dynamic client registration. No API key needs to be stored in the client configuration.

Connect with an API key

Server-side scripts and custom integrations can send a World Monitor key in the X-WorldMonitor-Key header:

curl -sS https://worldmonitor.app/mcp \
  -H "X-WorldMonitor-Key: wm_YOUR_KEY" \
  -H "Content-Type: application/json" \
  --data-binary '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "get_country_risk",
      "arguments": { "country_code": "AE" }
    }
  }'

Send a wm_... key through X-WorldMonitor-Key, not through the bearer-token header. Authorization: Bearer ... is reserved for OAuth access tokens.

Older or stdio-only clients can use the mcp-remote bridge:

{
  "mcpServers": {
    "worldmonitor": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://worldmonitor.app/mcp",
        "--header",
        "X-WorldMonitor-Key:wm_YOUR_KEY"
      ]
    }
  }
}

Public discovery and paid data access

The server deliberately separates public metadata from protected intelligence data.

The following discovery operations work without credentials:

• initialize
• tools/list
• prompts/list
• prompts/get
• resources/list
• resources/templates/list
• ping
• logging/setLevel

Data-bearing tools/call requests and protected resource reads require an active World Monitor Pro or API entitlement.

Current service limits:

• OAuth-connected contexts: 50 quota-consuming tool or protected-resource calls per UTC day.
• API-key contexts: 60 requests per minute per key; the MCP OAuth daily reservation does not apply.
• Anonymous discovery: 60 requests per minute per client IP.

See the live pricing and MCP documentation before building a workflow that depends on a specific allowance.

Smaller responses with JMESPath

Every tool accepts an optional jmespath argument. The server applies the expression before serializing the response, which can reduce context use substantially.

Example:

{
  "name": "get_market_data",
  "arguments": {
    "jmespath": "data.\"stocks-bootstrap\".quotes[?symbol=='AAPL' || symbol=='MSFT'].{symbol:symbol,price:price,change:change}"
  }
}

This is useful when an agent needs only a few fields from a broad market, conflict, or intelligence snapshot.

Example questions

• “Prepare a source-grounded world brief and identify the three most important escalation risks.”
• “Compare current country risk, macroeconomic pressure, and sanctions exposure for Egypt, Türkiye, and the UAE.”
• “Are maritime or aviation signals indicating disruption near the Strait of Hormuz?”
• “Which sectors are leading the US market, and how does that compare with commodities and risk sentiment?”
• “Show current conflict hotspots where news, military, and population-exposure signals converge.”
• “Build a morning intelligence brief for an energy trader and keep only decision-relevant fields.”
• “Check whether the latest result is fresh, stale, or degraded before drawing a conclusion.”
• “Find recent historical events that are similar to the current escalation pattern.”

Reliability and provenance

World Monitor data has different upstream cadences. A market quote, earthquake event, annual displacement series, and AI-generated brief do not have the same meaning or freshness budget. The MCP responses preserve that distinction instead of presenting every result as equally live.

Agents should:

1. Check cached_at and stale before making a time-sensitive claim.
2. Preserve source names, URLs, timestamps, and corroboration metadata in downstream answers.
3. Treat unavailable or degraded data as unknown, not as zero activity.
4. Use describe_tool when a compressed tool description does not provide enough detail.
5. Use JMESPath projection to keep multi-step workflows within a sensible context budget.

Compatibility

The remote endpoint works with MCP clients that support Streamable HTTP, including current versions of:

• Claude Desktop
• Claude on the web
• Cursor
• MCP Inspector
• Custom MCP clients and agent runtimes

Older stdio-only clients can connect through mcp-remote.

Links

• Homepage: worldmonitor.app
• Remote MCP endpoint: worldmonitor.app/mcp
• Five-minute quickstart: MCP Quickstart
• MCP server reference: MCP Overview
• Tool reference: MCP Tools Reference
• JMESPath guide: Response Projection
• Error catalog: MCP Error Catalog
• MCP Apps: Interactive MCP Apps
• Server card: Machine-readable MCP server card
• Status: status.worldmonitor.app
• Source code: github.com/koala73/worldmonitor
• Pricing and access: worldmonitor.app/pro

License

The World Monitor source code is available under AGPL-3.0-only. Commercial licensing is available for teams that need different terms. Data from upstream providers remains subject to each provider's applicable terms and attribution requirements.

---

World Monitor is built for analysts, researchers, journalists, risk teams, investors, developers, and AI agents that need a structured global operating picture with explicit sources and freshness—not an ungrounded answer generated from memory.