TrendRadar

TrendRadar exposes an MCP (Model Context Protocol) server that lets you interact with its real‑time news hotspot data using AI clients. You can run the MCP server locally or remotely, enabling natural language queries, trend analysis, and structured data access through compatible MCP clients. This guide walks you through practical usage, installation steps, and essential configuration so you can start querying and analyzing hot topics with ease.

How to use

Connect your MCP client to TrendRadar to perform interactive analyses and data queries. You can access the MCP via HTTP at a local gateway or run the MCP as a local process that your client communicates with directly. Typical tasks include asking the system for recent topics, comparing trends across platforms, and generating summarized insights from the stored news data. The MCP supports a suite of analysis tools that you can invoke from your client as standard MCP endpoints.

How to install

Prerequisites: you need a running environment for containers or a local runtime for the MCP, plus the ability to run HTTP or STDIO MCP endpoints. If you plan to use Docker, you will rely on the TrendRadar MCP images. If you prefer running the MCP locally, you will run the Python/UV based server as described below.

Step 1: Prepare a working directory for TrendRadar. You can clone the repository or download the project files into a new folder.

Step 2: Decide how you want to run the MCP server. There are two options: HTTP mode (remote) or STDIO mode (local). Use the option that matches your environment and deployment preference.

Step 3: Choose a specific command to start the MCP server and ensure it is accessible to your MCP client.

HTTP MCP server configuration

# Optional minimal example for HTTP mode (remote MCP server)
# This snippet shows an HTTP MCP endpoint configuration you would expose
# to your MCP client. Adjust according to your environment.
{
  "type": "http",
  "name": "trendradar_http",
  "url": "http://127.0.0.1:3333/mcp",
  "args": []
}

STDIO MCP server configuration

If you run the MCP server locally (STDIO mode), you configure a local command to start the server and pass any required arguments. The following example demonstrates the complete runtime command extracted from the reference flow.

{
  "type": "stdio",
  "name": "trendradar",
  "command": "uv",
  "args": [
    "run",
    "python",
    "-m",
    "mcp_server.server",
    "--transport",
    "http",
    "--port",
    "3333"
  ]
}

MCP client tools available

TrendRadar exposes a set of MCP endpoints used by clients to query and analyze data. The following tool endpoints are available within the MCP context:

• get_latest_news
• get_news_by_date
• get_trending_topics
• search_news
• search_related_news_history
• analyze_topic_trend
• analyze_data_insights
• analyze_sentiment
• find_similar_news
• generate_summary_report
• get_current_config
• get_system_status
• trigger_crawl

Available tools

get_latest_news

Fetches the latest news items from the configured sources or stored dataset.

get_news_by_date

Retrieves news items for a specific date range or date.

get_trending_topics

Returns currently trending topics based on predefined weighting and platform data.

search_news

Performs keyword-based searches across stored news data using MCP requests.

search_related_news_history

Finds historically related news items for given topics or keywords.

analyze_topic_trend

Analyzes the trend trajectory of a topic over time, including peak periods and lifecycle.

analyze_data_insights

Provides data-driven insights and cross-platform comparisons.

analyze_sentiment

Performs sentiment analysis on matched news items.

find_similar_news

Finds news articles similar to a given article or topic.

generate_summary_report

Generates a concise summary report for selected topics or dates.

get_current_config

Returns the current MCP server configuration and status.

get_system_status

Reports the health and status of the MCP service.

trigger_crawl

Manually trigger data crawling or data fetch for latest headlines.