🤖 Tutorial: Grok AI Integration & Sentiment Analysis with Claude

Home › Docs › Claude Tutorials › Tutorial: Grok AI Integration & Sentiment Analysis with Claude
Claude Tutorials

Tutorial: Grok AI Integration & Sentiment Analysis with Claude

You are my AI-powered X/Twitter intelligence analyst. I want to use XActions' Grok AI integration and sentiment analysis tools to analyze content, monitor brand reputation, and get AI-powered insights from X. Help me set up and use these advanced AI features.

Context

I'm using XActions (https://github.com/nirholas/XActions), an open-source X/Twitter toolkit. It includes MCP tools for:

  • Querying Grok AI (X's built-in AI) via x_grok_query and x_grok_summarize
  • Sentiment analysis via x_analyze_sentiment (built-in rules engine or LLM via OpenRouter)
  • Reputation monitoring via x_monitor_reputation and x_reputation_report

What I Need You To Do

Phase 1: Grok AI Queries

Use Grok (X's AI) directly through XActions:

  1. Ask Grok a question:

    "Ask Grok: What are the trending topics in AI today?"

    Tool: x_grok_query

    • query: The question or prompt for Grok

    Grok can answer questions about:

    • Current events and trending topics on X
    • Summarize discussions happening on X
    • Provide context on viral tweets
    • Answer general knowledge questions

  2. Summarize content with Grok:

    "Use Grok to summarize the discourse around 'AI regulation' on X"

    Tool: x_grok_summarize

    • Gets AI-powered summaries of conversations and topics

  3. Example Grok queries to try:

    • "What are people saying about [product] on X right now?"
    • "Summarize the debate about [topic] on X"
    • "What are the most viral topics in [industry] today?"
    • "Who are the most influential voices discussing [keyword]?"

Phase 2: Sentiment Analysis

Analyze the sentiment of any text using XActions' built-in analyzer:

  1. Analyze a single text:

    "Analyze the sentiment of this tweet: 'This product is absolutely amazing, best purchase I've ever made!'"

    Tool: x_analyze_sentiment

    • text: The text to analyze
    • Returns: score (-1 to 1), label (positive/neutral/negative), confidence, key words

  2. Batch analysis:

    "Analyze sentiment for these tweets: ['Great product!', 'Terrible experience', 'It was okay I guess']"
    • texts: Array of strings for batch analysis
    • Returns individual scores plus aggregate

  3. Analysis modes:

    • "rules" mode (default): Offline, zero dependencies, instant. Uses built-in word lists and linguistic rules
    • "llm" mode: Uses OpenRouter API for nuanced, context-aware analysis. Requires OPENROUTER_API_KEY env var
    "Analyze sentiment in LLM mode: 'I can't believe how bad the response time was'"

  4. Interpret results:

    • Score: -1.0 (very negative) → 0.0 (neutral) → +1.0 (very positive)
    • Label: "positive", "neutral", "negative"
    • Confidence: 0.0 to 1.0 (how certain the analysis is)
    • Keywords: Which words drove the sentiment score

  5. Practical workflows:

    Analyze mentions sentiment:

    Step 1: "Search for tweets mentioning @mycompany in the last 24 hours"
Step 2: "Now analyze the sentiment of each of these tweets"
Step 3: "Summarize: what percentage are positive vs negative?"

    Analyze competitor perception:

    Step 1: "Search tweets about CompetitorProduct"
Step 2: "Analyze sentiment batch for all these tweets"
Step 3: "Compare: how does sentiment about CompetitorProduct compare to our product?"

Phase 3: Reputation Monitoring

Set up continuous reputation monitoring:

  1. Start a reputation monitor:

    "Start monitoring reputation for @mycompany"

    Tool: x_monitor_reputation

    • action: "start"
    • target: "@mycompany" (username with @) or "product name" (keyword)
    • type: "mentions" (default), "keyword", or "replies"
    • interval: Polling interval in seconds (default: 900 = 15 min, minimum: 60)
    • webhookUrl: Optional URL for real-time alerts

    What it does:

    • Periodically scrapes mentions/keyword results
    • Runs sentiment analysis on each new mention
    • Computes rolling averages
    • Detects sentiment anomalies (sudden drops or spikes)
    • Sends webhook alerts for anomalies

  2. Configure webhook alerts:

    "Start monitoring @mycompany with webhook alerts to https://hooks.slack.com/xxx"

    This sends alerts when:

    • Sudden negative sentiment spike detected
    • Unusual mention volume detected
    • Sentiment drops below threshold

  3. Monitor types:

    • mentions: Track @mentions of a username
    • keyword: Track a brand name or product keyword
    • replies: Track replies to a specific user's tweets

  4. List active monitors:

    "List all my reputation monitors"
    • action: "list"
    • Shows: target, type, interval, status, poll count

  5. Check monitor status:

    "Show status for monitor ID xyz"
    • action: "status"
    • monitorId: The returned ID
    • Shows: latest data, sentiment trend, alert history

  6. Stop a monitor:

    "Stop reputation monitor xyz"
    • action: "stop"
    • monitorId: The ID to stop

Phase 4: Reputation Reports

Generate comprehensive reputation reports:

  1. Generate a report:

    "Generate a 7-day reputation report for @mycompany"

    Tool: x_reputation_report

    • username: Target username
    • period: "24h", "7d", "30d", or "all"
    • format: "json" or "markdown" (default: markdown)

  2. What the report includes:

    • Sentiment Distribution: % positive, neutral, negative
    • Top Positive Mentions: Best things people are saying
    • Top Negative Mentions: Complaints and criticism
    • Timeline Data: How sentiment changed over time
    • Keyword Frequency: Most common words in mentions
    • Alerts: Any anomalies or concerning trends
    • Volume Analysis: Mention frequency over time

  3. Report periods:

    • 24h: Real-time crisis check — are we being dragged right now?
    • 7d: Weekly health check — how was the week?
    • 30d: Monthly trend — where are we heading?
    • all: Complete historical view

  4. Example: Crisis detection workflow:

    Step 1: "Start monitoring @mycompany every 5 minutes" (interval: 300)
Step 2: "Generate a 24h reputation report"
Step 3: "Show me all negative mentions from the report"
Step 4: "Draft responses to the top 3 complaints"

Phase 5: Combined Intelligence Workflows

Brand Health Dashboard

1. "Start reputation monitor for @mycompany"
2. "Search recent tweets about our brand and analyze sentiment"
3. "Generate a 7-day reputation report"
4. "Compare our sentiment to @competitor's sentiment"
5. "Summarize: What should we focus on this week?"

Content Strategy from AI Insights

1. "Ask Grok what topics are trending in [my niche]"
2. "Search tweets about top 3 trending topics"
3. "Analyze which topics have the most positive sentiment"
4. "Recommend: What should I tweet about today?"

Customer Sentiment Tracking

1. "Monitor keywords: 'product name', 'brand name'"
2. "Generate daily sentiment reports"
3. "Flag any tweet with sentiment score below -0.5"
4. "Draft empathetic responses for negative mentions"

Competitor Intelligence with Sentiment

1. "Monitor mentions of @competitor1, @competitor2, @competitor3"
2. "Run sentiment analysis on each competitor's mentions"
3. "Compare: sentiment distribution across competitors"
4. "Identify: what complaints do their users have that we solve?"

Phase 6: OpenRouter LLM Integration

For more nuanced sentiment analysis, connect an LLM:

  1. Setup:

    • Get an API key from OpenRouter
    • Set OPENROUTER_API_KEY in your environment

  2. LLM vs Rules comparison:

    Feature Rules Mode LLM Mode
    Speed Instant 1-3 seconds
    Cost Free Per-token
    Nuance Good Excellent
    Sarcasm Poor Good
    Context None Full
    Offline Yes No

  3. When to use LLM mode:

    • Analyzing sarcastic or ironic tweets
    • Need nuanced emotional categorization
    • Detecting subtle negativity or passive-aggression
    • Multi-language sentiment analysis

  4. Example:

    "Analyze in LLM mode: 'Oh great, another update that breaks everything. Thanks SO much!'"
    • Rules mode might score this as positive (words: "great", "thanks")
    • LLM mode correctly identifies sarcasm → negative

My AI Analysis Goals

(Replace before pasting)

  • Brand/account to monitor: @myhandle
  • Competitors to track: @comp1, @comp2
  • Keywords to monitor: keyword1, keyword2
  • Alert webhook URL: (optional)
  • OpenRouter API key available: Yes/No

Start with Phase 1 — let's try a Grok query first, then set up sentiment analysis on my mentions.

⚡ Explore XActions

100% free and open-source. No API keys, no fees, no signup.

Browse All Documentation