Overview

The Audiense Insights MCP Server implements the Model Context Protocol (MCP) adapter that connects MCP-compatible clients—such as Claude—to an Audiense Insights account. Once running, the server exposes a set of MCP tools that let an LLM client retrieve audience analytics, influencer comparisons, content engagement, baselines and metadata from Audiense reports so the model can reason over real marketing data in-context.

Note: this repository historically provided a local MCP server. Audiense has migrated the Insights MCP to a remote model; this document describes the original server and how it was configured. If you plan production usage, check with Audiense support for the current remote MCP options ([email protected]).

Features

Authenticate to Audiense Insights via client credentials and serve data to MCP clients
Fetch lists of intelligence reports and detailed report metadata
Retrieve aggregated audience insights: demographics, psychographics, behaviors
Compare audience influencers against baseline audiences with affinity and uniqueness scores
Retrieve audience content engagement: popular posts, domains, emojis, hashtags, links, media and wordclouds
Optional enrichment of influencer data using an X/Twitter bearer token
Simple integration pattern for Claude Desktop (or any MCP-capable client)

Installation / Configuration

Prerequisites:

Node.js v18+
MCP-compatible client (e.g., Claude Desktop)
Audiense Insights account with API client ID and client secret
Optional: X/Twitter API Bearer Token for influencer enrichment

Install and run the MCP server locally (example using npx):

# Run without installing globally
npx -y mcp-audiense-insights

Set required environment variables (POSIX shell example):

export AUDIENSE_CLIENT_ID="your_client_id"
export AUDIENSE_CLIENT_SECRET="your_client_secret"
# Optional
export TWITTER_BEARER_TOKEN="your_twitter_bearer_token"

Windows (PowerShell):

$env:AUDIENSE_CLIENT_ID = "your_client_id"
$env:AUDIENSE_CLIENT_SECRET = "your_client_secret"
# Optional
$env:TWITTER_BEARER_TOKEN = "your_twitter_bearer_token"

Configuring Claude Desktop to launch the MCP server automatically:

MacOS: edit ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: edit %AppData%\Claude\claude_desktop_config.json

Add an MCP server entry, for example:

"mcpServers": {
  "audiense-insights": {
    "command": "npx",
    "args": ["-y", "mcp-audiense-insights"],
    "env": {
      "AUDIENSE_CLIENT_ID": "your_client_id_here",
      "AUDIENSE_CLIENT_SECRET": "your_client_secret_here",
      "TWITTER_BEARER_TOKEN": "your_token_here"
    }
  }
}

Save and restart Claude Desktop so it can discover the local MCP tools.

Available Tools

The MCP server exposes a set of JSON-RPC-like tools (descriptions simplified):

Tool name	Purpose	Key parameters
get-reports	List Audiense intelligence reports for the authenticated user	none
get-report-info	Get detailed metadata for a specific report	report_id (string)
get-audience-insights	Return aggregated insights for an audience	audience_insights_id (string), insights (array, optional)
get-baselines	List baseline audiences (optionally by country)	country (ISO code, optional)
get-categories	Fetch available affinity/categories for comparisons	none
compare-audience-influencers	Compare audience influencers vs baseline	audience_influencers_id, baseline_audience_influencers_id, pagination & filters
get-audience-content	Retrieve content engagement (liked/shared/influential)	audience_content_id (string)
report-summary	Generate a natural-language summary of a report	report_id (string)

Each tool returns structured JSON. Several endpoints support pagination (cursor, count) and filters like followers_min/max, entity_type, bio_keyword, categories and countries.

Use Cases

Quickly surface audience demographics to an LLM assistant

Action: call get-audience-insights with an audience_insights_id
Outcome: Claude can produce a succinct summary (“audience is 65% male, predominant age 25–34, top countries: US, UK”) and use that in downstream content prompts.

Identify unique influencers for a campaign

Action: call compare-audience-influencers with the audience and baseline IDs, filter by followers_min and categories
Outcome: obtain influencers with affinity, baseline affinity and uniqueness scores; export top candidates to a CSV for outreach.

Inform content strategy with engagement signals

Action: call get-audience-content to retrieve topHashtags, topDomains, topMedia, and wordcloud
Outcome: generate topic and format recommendations (e.g., short-form videos and posts using hashtags X, Y) tailored to what resonates with the audience.

Example filter use (conceptual):

{
  "tool": "compare-audience-influencers",
  "params": {
    "audience_influencers_id": "aud-123",
    "baseline_audience_influencers_id": "global",
    "followers_min": 10000,
    "categories": ["technology","gaming"],
    "count": 100
  }
}

Notes and Best Practices

Store credentials securely (avoid committing client secrets to source control).
Respect Audiense API rate limits; use pagination and sensible count limits (default count is 200 for influencer lists).
The baseline logic: if one country >50% of the audience it becomes the baseline; otherwise the global baseline is used.
The repository historically provided a local adapter; confirm whether a maintained remote MCP is available for production usage via Audiense.

For help or to confirm the current supported integration, contact Audiense support ([email protected]).

Audiense Insights MCP Server for Claude Integration