SentimentAlpha.ai

Live Metrics

Six Oracle Tools

One MCP server. Six specialized oracles. Five paid at $0.01/query via x402. One free FinOps advisor for any x402 service.

All six tools share the same dual-layer cache and MCP manifest. Five paid via x402. Spend Guard is free. One connection. Unlimited intelligence.

API Endpoints

Oracle Tools (x402 paid)

Free FinOps Oracle

Free Previews (agent eval)

Discovery & Health

Quick Start for Agents

1. Discover: GET /v1/mcp-manifest — 6 tools auto-registered

2. Try free: POST /v1/spend-guard-alpha (free), or any preview endpoint (e.g. /narrative-alpha/preview)

3. Pay & query: Send 0.01 USDC on Base — wallet provided in 402 response

4. Attach receipt: POST /v1/narrative-alpha, /manipulation-alpha, /convergence-alpha, /reputation-alpha, or /alert-alpha with X-Payment header

Try Every Tool — Free Previews

Paste any of these into your terminal. No API key, no signup, no payment.

1. Narrative Alpha — What's the sentiment?

curl -s -X POST https://sentimentalpha.ai/v1/narrative-alpha/preview \
  -H "Content-Type: application/json" \
  -d '{"query": "$BTC"}' | python3 -m json.tool

2. Manipulation Risk — Is it real or coordinated?

curl -s -X POST https://sentimentalpha.ai/v1/manipulation-alpha/preview \
  -H "Content-Type: application/json" \
  -d '{"query": "$PEPE"}' | python3 -m json.tool

3. Convergence Alpha — Are all signals aligned?

curl -s -X POST https://sentimentalpha.ai/v1/convergence-alpha/preview \
  -H "Content-Type: application/json" \
  -d '{"query": "$SOL"}' | python3 -m json.tool

4. Spend Guard — Can I afford this? (free)

curl -s -X POST https://sentimentalpha.ai/v1/spend-guard-alpha \
  -H "Content-Type: application/json" \
  -d '{"query": "Running 50 queries/hour across 3 x402 oracles at $0.01 each"}' | python3 -m json.tool

5. Reputation Alpha — Can I trust this entity?

curl -s -X POST https://sentimentalpha.ai/v1/reputation-alpha/preview \
  -H "Content-Type: application/json" \
  -d '{"query": "vitalik.eth", "entity_type": "wallet"}' | python3 -m json.tool

6. Alert Alpha — Watch my portfolio for me

curl -s -X POST https://sentimentalpha.ai/v1/alert-alpha/preview \
  -H "Content-Type: application/json" \
  -d '{"query": "portfolio watch", "watchlist": [
    {"ticker": "$BTC", "condition": "sentiment_below", "threshold": 0.3},
    {"ticker": "$ETH", "condition": "manipulation_above", "threshold": 70},
    {"ticker": "$SOL", "condition": "narrative_velocity_above", "threshold": 80}
  ]}' | python3 -m json.tool

Every tool returns structured JSON designed for programmatic agent consumption. Previews are redacted — pay $0.01 USDC via x402 for full data.

Python (3 lines)

import requests
r = requests.post("https://sentimentalpha.ai/v1/narrative-alpha/preview",
                   json={"query": "$BTC"})
print(r.json()["actionableInsight"])

MCP / Claude Desktop

Add to claude_desktop_config.json:

{ "mcpServers": { "sentimentalpha": {
    "command": "npx",
    "args": ["-y", "mcp-remote", "https://sentimentalpha.ai/v1/mcp-manifest"]
}}}

Full integration examples (LangChain, CrewAI, Claude): github.com/tharmer/x-narrative-oracle/examples

Example Response Schema

Every paid query returns this structured JSON — designed for programmatic consumption by autonomous agents, not human dashboards.

{
  "sentiment": {
    "score": 0.62,
    "label": "greed",
    "confidence": 0.81
  },
  "narrativeVelocity": {
    "score": 74,
    "trend": "accelerating",
    "hourlyMentions": 2840
  },
  "topNarratives": [
    {
      "theme": "ETF inflow momentum",
      "sentiment": "bullish",
      "strength": 88,
      "sampleTake": "BlackRock BTC ETF just had its biggest inflow day since launch"
    },
    {
      "theme": "Whale accumulation on-chain",
      "sentiment": "bullish",
      "strength": 72,
      "sampleTake": "Top 100 wallets added 12k BTC this week per Glassnode"
    }
  ],
  "keyInfluencers": [
    { "handle": "@sabortoalpha", "stance": "bullish, targeting 120k", "reach": "high" }
  ],
  "contrarian": {
    "hasSignal": true,
    "consensusDivergence": 82,
    "description": "Extreme bullish consensus (82/100) — historically precedes 5-15% pullbacks within 7 days"
  },
  "manipulationRisk": {
    "score": 18,
    "flags": ["low_organic_ratio"]
  },
  "relatedTickers": ["$BTC", "$IBIT", "$MSTR", "$COIN"],
  "actionableInsight": "Strong bullish momentum but consensus divergence at 82 suggests profit-taking window. Smart money historically fades this level of X euphoria.",
  "dataQuality": {
    "postCount": 4200,
    "timeRange": "last 4 hours",
    "freshness": "real-time",
    "sourceModel": "grok-3-fast-x_search"
  }
}

Why Not Just Ask Grok Directly?

Consistent schema: Raw LLM output varies every call. We enforce a strict JSON contract so your agent never breaks on unexpected formats.

Manipulation detection: We flag bot-driven narratives, astroturf patterns, and coordinated amplification — something a raw Grok prompt won't do.

Contrarian scoring: Our consensus divergence metric quantifies how one-sided the crowd is, surfacing reversal risk that raw sentiment misses.

Sub-400ms cached responses: Popular queries hit our cache layer, returning in <400ms vs 3-8s for a fresh Grok call.

Cross-asset ticker extraction: Automatically identifies related assets mentioned in the narrative for multi-leg trade ideation.