NERAI Risk Intelligence — MCP Server
The geopolitical grounding layer for AI agents. Your procurement, trading or compliance agent knows your systems; it does not know the world. NERAI is the world, as a tool call — 60 countries, calibrated forecasts, maritime chokepoints, sanctions and trade-control data, every answer source-linked and dated.
Remote MCP server (Streamable HTTP) — no install, no local process:
https://nerai-mcp.neraicorp.workers.dev/mcp
Quick start
Get a free key instantly (no card, no call):
curl -X POST https://nerai-mcp.neraicorp.workers.dev/register \
-H 'content-type: application/json' -d '{"email":"you@company.com"}'
Connect from Claude Code (any MCP client works the same way):
claude mcp add nerai --transport http https://nerai-mcp.neraicorp.workers.dev/mcp \
--header "Authorization: Bearer <your-key>"
Clients without a header field (e.g. Claude.ai custom connectors):
https://nerai-mcp.neraicorp.workers.dev/mcp/<your-key>
What your agent gets
16 tools across five layers:
| Layer | Tools |
|---|---|
| Country risk | list_countries · get_country_risk · compare_countries · get_causal_drivers |
| Foresight | get_forecast_desk (re-scored every 2 days, Brier-tracked) · get_early_warning |
| Supply chain | get_supply_chain_risk · get_commodities |
| Maritime | get_maritime_corridors · get_live_chokepoint_traffic (live AIS) · get_circumvention_corridors |
| Trade controls | get_sanctions_overview · get_tradecontrol_changes · check_commodity_controls · find_enforcement_precedent · get_transaction_intelligence_status |
Ask it something real
Once connected, try these in your agent:
"Which of these supplier countries deteriorated this month: Turkey, Kazakhstan, Egypt?" "What is the live vessel count in the Strait of Hormuz versus baseline?" "What changed in the UK and EU trade-control registers this week? Cite the legal instruments." "Compare the risk trajectory of Poland and Romania over the next 13 weeks."
Why this one
- Validation is published, not promised. Out-of-sample MASE 0.85–0.95 and 61–67% directional accuracy across 4–52 week horizons. Ask your data vendor for their Brier score — we publish ours.
- Agent-safe by contract. Every response carries
_meta:as_of, source, refresh cadence and limitations. An agent cannot ask a follow-up question, so the answer states its own age and limits. That block is also your audit trail. - Enforcement precedents are human-reviewed. Client-facing tools return only analyst-signed cases, each linking the authority's own document.
Endpoints
GET /pricing— machine-readable tier table (free / pro / enterprise)GET /llms.txt— agent-readable capability sheetGET /openapi.json— transport spec with the full tool cataloguePOST /register— instant free key
Recipes
Copy-paste agent patterns in recipes/: a procurement agent that
checks the world before re-ordering, a trading agent grounded in world state,
and a compliance agent that cites official sources and leaves signing to a human.
Pricing
Free (5 countries, 100 calls/mo) · Pro €490/mo or €4,900/yr (all 60 countries, 100k calls) · Enterprise from €15,000/yr (HS-level commodity checks, reviewed precedent corpus, SLA). Details: neraicorp.com/ai-agents.html
Not legal or investment advice. Built and operated by NERAI BV, Brussels. Contact: kagan@neraicorp.com