X Intelligence MCP Server

Twitter/X content discovery via public search sources, exposed as MCP tools for AI agents. Free demo tool for exploring social media mentions, brand sentiment, and buying-intent signals.

Important Limitations

This is a free demo tool with significant limitations.

Feature	Status
Real-time data	No - uses public/indexed sources (may be hours/days old)
Engagement metrics	No - likes, retweets, replies are returned as 0 from public sources
Full tweet text	Partial - RSS titles and search-engine snippets only
Rate limits	Subject to public source rate limits (Nitter / DuckDuckGo)
Reliability	Best-effort, may return no results or fall back to sample data

For production use cases requiring real-time Twitter data with engagement metrics, you need:

• Twitter API Pro - developer.twitter.com
• Or a licensed data provider (Brandwatch, Sprinklr, etc.)

How It Works

The server collects tweet-like content from public sources, in this order:

1. Nitter RSS feeds, then Nitter HTML (open-source Twitter frontends; many instances are offline)
2. Twitter syndication API (limited availability)
3. DuckDuckGo site:twitter.com / site:x.com search snippets
4. Apify Twitter scraper fallback (apidojo/twitter-scraper-lite) — only if an Apify token is available and the public sources return nothing
5. Sample/demo data — if every source above fails, a small set of clearly illustrative placeholder tweets is returned so the tool still responds

Optional OpenAI analysis then runs sentiment/insight extraction on the collected items. Tools are exposed via Apify's MCP infrastructure (mcp.apify.com); each tool call runs the actor and returns its result.

Input Parameters

Field	Type	Required	Default	Description
apifyToken	string (secret)	See note	—	Apify API token used for the Apify-scraper fallback path. On the Apify platform the platform token (APIFY_TOKEN) is used automatically. Required for every tool except get_stats (the run stops with a graceful missing_apify_token error otherwise). No X/Twitter login is required.
openaiApiKey	string (secret)	No	—	OpenAI API key for AI sentiment/insight analysis. Without it you get raw results only. Billed to your key.
openaiModel	string (enum)	No	gpt-4o-mini	Model for analysis. One of gpt-4o-mini, gpt-4o, gpt-4-turbo, gpt-3.5-turbo.
scraperTimeout	integer	No	120	Max seconds to wait for the Apify scraper fallback (30–300).
toolCall	object	No	x_search demo call	The MCP tool to execute: { "name": <tool>, "arguments": { ... } }. Leave empty to only return server info.

Tools

Six tools are exposed. Each toolCall.arguments object accepts the parameters below.

x_search

Search Twitter/X content matching a query.

• query (string, required)
• max_tweets (integer, default 20)
• analyze (boolean, default true) — runs OpenAI sentiment analysis when an OpenAI key is set; otherwise the analysis field reports that no key is configured.

{
  "name": "x_search",
  "arguments": { "query": "your search term", "max_tweets": 10, "analyze": false }
}

x_monitor

Collect brand mentions and run AI sentiment/perception analysis.

• brand (string, required)
• max_tweets (integer, default 50)
• sentiment_filter (string, default "any") — accepted for forward compatibility but not currently applied to filter results.

{
  "name": "x_monitor",
  "arguments": { "brand": "Brand Name", "max_tweets": 20 }
}

x_intent

Find buying-intent signals (recommendations sought, frustration, comparisons, switching).

• product_category (string, required)
• intent_types (array of strings, default ["seeking_recommendation", "frustrated", "comparing", "switching"])
• max_tweets (integer, default 30)

{
  "name": "x_intent",
  "arguments": { "product_category": "CRM software", "max_tweets": 15 }
}

x_compete

Compare sentiment and share of voice across 2–5 brands.

• brands (array of strings, required, 2–5 entries)
• tweets_per_brand (integer, default 30)

{
  "name": "x_compete",
  "arguments": { "brands": ["Brand A", "Brand B"], "tweets_per_brand": 10 }
}

x_profile

Return recent content for a specific Twitter/X user.

• username (string, required, without @)
• include_tweets (boolean, default true)
• max_tweets (integer, default 20)

Note: detailed profile metadata (bio, follower/following counts, verified status) is only populated when the Apify scraper path runs; via the free public path those fields are usually empty and the tool returns recent tweets/content only.

{
  "name": "x_profile",
  "arguments": { "username": "username", "max_tweets": 10 }
}

get_stats

Return server version, usage counters, and whether AI analysis is enabled. Takes no arguments and requires no Apify token.

Setup

1. Optional: OpenAI Key for AI Analysis

Without an OpenAI key you get raw collected results only. With a key you get sentiment analysis and insights. Get a key at platform.openai.com/api-keys.

2. Connect to an MCP Client

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "x-intelligence": {
      "url": "https://mcp.apify.com?actors=YOUR_USERNAME/x-intelligence-mcp"
    }
  }
}

3. Use It

Ask Claude: "Search Twitter for mentions of Acme Corp"

Example Output

Results are pushed to the run's dataset. A tool_result record looks like:

{
  "record_type": "tool_result",
  "tool": "x_search",
  "success": true,
  "tweet_count": 5,
  "query": "Acme Corp",
  "tweets": [
    {
      "text": "Just tried Acme Corp's new product...",
      "author": "someuser",
      "likes": 0,
      "retweets": 0,
      "url": "https://twitter.com/someuser/status/123"
    }
  ],
  "result": { "...": "full tool response, incl. AI analysis when enabled" }
}

The run also pushes an mcp_server_info record (server name, version, tool definitions, capabilities) and a final session_stats record. Engagement metrics (likes, retweets) are not available through the public sources and are returned as 0.

Pricing

• Running the actor over public sources is effectively free (Apify platform compute usage applies).
• Optional OpenAI analysis is billed to your own OpenAI API key (a few cents per analysis).
• The Apify Twitter-scraper fallback (apidojo/twitter-scraper-lite), if triggered, incurs that third-party actor's own usage costs.

Costs above are indicative; check current provider pricing.

Limitations

• Not real-time. Data comes from public/indexed sources and can be hours to days old.
• No engagement metrics. Likes, retweets, and replies from public sources are 0.
• Sample-data fallback. If Nitter, syndication, DuckDuckGo, and the Apify scraper all fail, the tools return a small set of clearly illustrative placeholder tweets — not real data.
• Apify token required for most tools. Every tool except get_stats needs an Apify token; on the Apify platform this is provided automatically, but other environments must supply apifyToken or the run stops with a graceful missing_apify_token error. (No X/Twitter credentials are needed — only public sources are used.)
• AI analysis needs an OpenAI key. Without openaiApiKey, analysis fields report that no key is configured and you receive raw results only.
• Rate limits and availability. Public sources (Nitter instances, DuckDuckGo) are frequently offline or rate-limited; results are best-effort and may be empty.
• Limited profile data. x_profile bio/follower fields are only populated on the Apify-scraper path and are usually empty via the free public path.
• sentiment_filter is not applied. x_monitor accepts the argument but does not currently filter results by it.
• Standby / MCP mode. The actor is invoked per tool call through Apify's MCP bridge; it is not a persistent long-running server.

Support

This is a free demo tool provided as-is. For production needs, consider official Twitter/X API access or a licensed data provider.