Skip to content
← Blog
5 min read

Fact-Check X Posts and Get Social Sentiment — New MCP Bridge

Your AI agent can now search X, analyze social sentiment, and verify claims in trending posts. Two sentiment layers with divergence detection. One npm install.

X is the fastest public signal on Earth. When earnings drop, when a CEO tweets, when a rumor starts — it shows up on X before anywhere else. But raw social data is noisy. Most of it is opinion, hype, or outright wrong.

Today we're shipping @veroq/x-bridge — an MCP server that lets AI agents search X, analyze social sentiment, and fact-check the most-shared claims in real time. Your agent gets the speed of social media with the accuracy of verified intelligence.

The Problem: Social Speed vs. Accuracy

AI agents that consume X data face a dilemma. Social media is fast but unreliable. News is reliable but slow. By the time Reuters confirms a rumor, the market already moved.

The bridge solves this by combining both: search X for the signal, then verify the claims via VeroQ Shield before acting on them. Speed of social, accuracy of verification.

5 Tools, One Install

bash
npm install @veroq/x-bridge
ToolWhat it does
x_verify_postFact-check claims in a specific X post via Shield
x_topic_sentimentSocial sentiment from X + VeroQ news sentiment
x_verify_trendingSearch + verify top posts by engagement
x_entity_pulseFull intel: social + news + briefs + entity graph
x_search_verifiedSearch X with auto-verification of top result

Two Sentiment Layers

The bridge provides two independent sentiment signals that your agent can compare:

Layer 1: X Social Sentiment (real-time)

Lightweight keyword analysis across all searched posts. Detects bullish/bearish signals, aggregates into distribution (positive/negative/neutral/mixed), and surfaces the top signal words. Zero API cost — runs locally on every post.

Layer 2: VeroQ News Sentiment (verified)

Entity-level sentiment extracted from verified news briefs. 7-day, 30-day, or 90-day lookback with trend direction (improving/declining/stable). Based on actual reporting, not social noise.

Divergence Detection

When X social says “bullish” but news says “declining”, the bridge flags the divergence. Social often leads news by 12–48 hours — a divergence can be an early signal before traditional media catches up.

Example: Entity Pulse

Ask your agent: “Give me the full picture on NVIDIA.”

The x_entity_pulse tool fires 4 API calls in parallel and returns the full picture. Post content is visible to the agent locally via your own X bearer token — the hosted API endpoints return only aggregated analytics, never raw post text.

text
## Entity Pulse: NVIDIA

### X Social Sentiment (real-time)
Posts: 50 | Sentiment: positive (0.42)
Distribution: 18+ 6- 22~ 4m
Bullish: growth, record, strong, rally
Top post: @chipanalyst: "NVIDIA data center revenue..." (2.3K likes)

### News Sentiment (7d)
Trend: stable
Coverage: 12 articles — 4 positive, 2 negative, 6 neutral
Aligned: Social and news sentiment both positive

### Latest Verified News
- NVIDIA Reports Record Q1 Data Center Revenue (94% confidence)
- Blackwell Architecture Drives AI Inference Demand (87% confidence)

### Related Entities
- AMD — 22 co-mentions, sentiment: neutral
- OpenAI — 24 co-mentions, sentiment: positive
- Google — 21 co-mentions, sentiment: neutral

Verify Before You Act

The most powerful pattern: search X for what people are saying, then verify the top claims before your agent acts on them.

text
## Verified Trending: "Tesla earnings"
Searched: 50 posts | Verifying top: 3

### @marketwatch_bot (1.2K likes, 850K views)
> "Tesla Q1 revenue hit $25.7B, beating estimates by 12%"

Trust Score: 0.87 | Verdict: partially_verified
  [+] Tesla Q1 revenue was $25.7 billion — supported (91%)
  [x] Revenue beat estimates by 12% — contradicted (85%)
      Correction: Revenue beat estimates by 4.2%, not 12%

### Social Sentiment (50 posts)
Overall: positive (0.38) | 16+ 8- 22~ 4m

The post was mostly right — but the “12% beat” was wrong. Without verification, your agent would have acted on inflated expectations.

Setup: Claude Desktop

json
{
  "mcpServers": {
    "veroq-x-bridge": {
      "command": "npx",
      "args": ["@veroq/x-bridge"],
      "env": {
        "VEROQ_API_KEY": "your_key",
        "X_BEARER_TOKEN": "your_token"
      }
    }
  }
}

Each user brings their own API keys. X API is pay-per-use (no monthly subscription). VeroQ is free up to 1,000 credits/month.

Hosted API: No X Key Needed

Don't want to manage X API credentials? Use the hosted social sentiment endpoints directly through the VeroQ API:

bash
# Social sentiment for any entity
curl https://api.veroq.ai/api/v1/social/sentiment/NVIDIA \
  -H "Authorization: Bearer YOUR_VEROQ_KEY"

# Social sentiment for a ticker
curl https://api.veroq.ai/api/v1/ticker/NVDA/social \
  -H "Authorization: Bearer YOUR_VEROQ_KEY"

# Trending tickers on social media
curl https://api.veroq.ai/api/v1/social/trending \
  -H "Authorization: Bearer YOUR_VEROQ_KEY"

These endpoints combine Reddit and X data, run LLM-powered sentiment analysis, and return structured results. Available through the API, MCP, and all SDKs. Hosted social endpoints require a paid plan (Builder and above) because X API lookups are pay-per-use — self-host via @veroq/x-bridge with your own X key on the free tier.

What's Next

Social data is the fastest public signal available. Combined with verified intelligence and claim-level fact-checking, your agent can move at the speed of social media without the risk of acting on bad information.