# Leadership Under Uncertainty — Strategic Cognition MCP

> Independent research project on AI, judgment, and strategic cognition. Publishes open benchmarks and ontologies for evaluating AI systems on strategic-analytic tasks. A reference MCP server is available for research use.

## What is this?

Leadership Under Uncertainty is an independent research project led by Christopher Berzins. It publishes:

- **Strategic Cognition Benchmark (SCB)** — a two-layer benchmark evaluating AI systems against ICD 203 analytic standards
- **Strategic Cognition Ontology (SCO)** — a reference ontology for strategic reasoning
- **Reference MCP server** — an implementation of the SCB/SCO standards, available for research and evaluation use

All research artifacts are open under permissive licenses (CC-BY 4.0 for papers, Apache 2.0 for code, CC-BY-SA 4.0 for the benchmark dataset).

## How to Authenticate

Pass an API key as a Bearer token. Keys use the `nb_mcp_` prefix (a legacy identifier preserved for protocol compatibility).

```
POST https://leadershipunderuncertainty.org/api/mcp
Content-Type: application/json
Authorization: Bearer nb_mcp_xxxxxxxxxxxxxxxx
```

Manage keys at: https://leadershipunderuncertainty.org/settings/api-keys

## Available Tools

### Briefing & Context
- `getBrief(dateKey?)` — Generate a structured strategic intelligence brief. Cost: $0.01-$0.10, Latency: 2-10s
- `getDailySections(sections[], dateKey?)` — Get specific sections of a daily brief. Cost: $0.01-$0.10, Latency: 2-10s
- `getCachedContext()` — Get user context (goals, views, SLS, recent signals). Cost: $0.001, Latency: 50-200ms

### Evidence & Signals
- `queryEvidence(opts?)` — List recent signals with optional filters (priority, domain, date). Cost: $0.001, Latency: 50-500ms
- `searchEvidence(query, limit?)` — Full-text search across evidence. Cost: $0.005, Latency: 200-2000ms
- `getSignal(id)` — Get full details for a specific signal. Cost: $0.001, Latency: 50-200ms

### Strategic Graph
- `getActors(opts?)` — List actors (people, organizations) in the strategic landscape. Cost: $0.001, Latency: 50-300ms
- `getStorylines(opts?)` — List active storylines/narratives being tracked. Cost: $0.001, Latency: 50-300ms
- `getAssumptions(opts?)` — List assumptions in the user's mental model. Cost: $0.001, Latency: 50-300ms
- `getPredictions(opts?)` — List tracked predictions with probabilities. Cost: $0.001, Latency: 50-300ms
- `getGraphEdges(entityId?)` — Get relationships between entities in the strategic graph. Cost: $0.001, Latency: 50-300ms
- `getGoals()` — Get the user's strategic goals. Cost: $0.001, Latency: 50-200ms

### Analysis (LLM-powered)
- `forecast(question, opts?)` — Generate a calibrated probability forecast. Modes: fast (~2s), superforecast (~30s), adw-delphi (~3-5min). Cost: $0.005-$0.40, Latency: 2-60s
- `assessRisk(topic, opts?)` — Multi-dimensional risk assessment. Cost: $0.01-$0.15, Latency: 3-20s
- `analyzeScenarios(question, opts?)` — Generate 2-5 weighted scenarios with probabilities. Cost: $0.01-$0.20, Latency: 5-30s
- `factCheck(claim, opts?)` — Fact-check a claim with structured evidence. Cost: $0.005-$0.10, Latency: 2-15s
- `contextualize(event, opts?)` — Contextualize an event against the user's strategic frame. Cost: $0.01-$0.15, Latency: 3-20s

### Mutations
- `createPrediction(question, probability, opts?)` — Record a new prediction. Cost: $0.001, Latency: 100-500ms
- `updatePrediction(id, updates)` — Update probability or rationale of a prediction. Cost: $0.001, Latency: 100-500ms

### Ontology & Self-Model
- `getOntology(opts?)` — Get the full SCO-compliant ontology (goals, views, actors, assumptions). Cost: $0.001-$0.01, Latency: 100-1000ms
- `getViews(opts?)` — Get strategic views with optional drivers. Cost: $0.001, Latency: 50-300ms
- `getSLS()` — Get the Strategic Living Script (narrative + key themes). Cost: $0.001, Latency: 50-200ms
- `getMentalModel()` — Get the user's mental model (biases, blind spots, strengths). Cost: $0.001, Latency: 50-200ms

## Example Usage Patterns

### Quick risk check
```json
{
  "jsonrpc": "2.0", "id": 1,
  "method": "tools/call",
  "params": {
    "name": "northbrief_risk",
    "arguments": { "topic": "EU AI Act compliance" }
  }
}
```

### Calibrated forecast
```json
{
  "jsonrpc": "2.0", "id": 1,
  "method": "tools/call",
  "params": {
    "name": "northbrief_forecast",
    "arguments": {
      "question": "Will the Fed cut rates before Q3 2026?",
      "mode": "superforecast"
    }
  }
}
```

### Get recent high-priority signals
```json
{
  "jsonrpc": "2.0", "id": 1,
  "method": "tools/call",
  "params": {
    "name": "northbrief_signals",
    "arguments": { "action": "list", "filter": "act", "limit": 10 }
  }
}
```

Note: the `northbrief_` tool-name prefix is a legacy protocol identifier preserved for client compatibility. Research publications, papers, and the public identity of this project are all under Leadership Under Uncertainty.

## Discovery Files

- **Agent Discovery:** https://leadershipunderuncertainty.org/.well-known/agents.md
- **A2A Agent Card:** https://leadershipunderuncertainty.org/.well-known/agent-card.json (canonical; the legacy /.well-known/agent.json path is also served for backward compatibility)
- **Capability Map (JSON):** https://leadershipunderuncertainty.org/api/sdk/capabilities
- **TypeScript Declarations:** https://leadershipunderuncertainty.org/api/sdk/types
- **OpenAPI Spec:** https://leadershipunderuncertainty.org/openapi.json
- **This file:** https://leadershipunderuncertainty.org/llms.txt

## Rate Limits

| Tier | Daily limit | Monthly budget |
|---|---|---|
| Anonymous | 5 | — |
| Registered | 50 | $5 |
| Verified | 200 | $25 |
| Research partner | Custom | Custom |

## Limitations

- Forecasts are probabilistic estimates, not certainties
- adw-delphi mode takes 3-5 minutes; use fast mode for quick checks
- Rate limits apply per API key tier
- Signal coverage focuses on geopolitics, markets, technology, and energy
- The MCP server reflects structured analytical reasoning, not real-time news feeds
- This is a research project, not a commercial product

## About

Leadership Under Uncertainty is an independent research project investigating AI judgment and strategic cognition. The published artifacts (SCB, SCO, reference MCP server) are open under permissive licenses and intended for research use and AI evaluation communities.

Website: https://leadershipunderuncertainty.org