AIMentionTracker

What AIMentionTracker MCP server does

The AIMentionTracker MCP server lets an MCP client read answer-engine optimisation data for tracked brands. Its measurements are based on answers generated for the workspace’s questions, including whether an answer mentioned or cited a brand. The supported engines are ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews.

The data covers brand visibility by engine, share of voice among tracked brands, classified sentiment, competitors named instead of the tracked brand, stored answers, citation details, source domains, alerts, and ranked recommendations. The service is intended for analysis and reporting; the tools are read-only and do not modify billing, team membership, tracked content, or workspace configuration.

How it works

A workspace API key is scoped to one organisation and determines which brands and questions the client can read. Requests are authenticated with that key and return data from the AIMentionTracker API. The hosted deployment creates a stateless server instance for each request and does not retain secrets, database credentials, vendor credentials, or service accounts.

Tool order matters when interpreting results. Start with get_me to identify the key, plan, and reporting thresholds. Call get_current_date before reasoning about reporting windows, then use list_brands and list_questions to confirm the available brands and the question coverage behind a measurement.

Measurements include interpretation safeguards. A null rate indicates that too few answers exist, not zero visibility. Engines are reported separately rather than averaged. Share of voice includes the number of brands counted, while sentiment uses the classified-answer denominator because some mentions cannot be classified reliably. An empty alert result does not necessarily mean that no change occurred; the response explains its meaning.

Setup and configuration

Create an API key through the AIMentionTracker API-key page. For hosted use, configure an MCP client with the streamable HTTP endpoint at https://mcp.aimentiontracker.ai and send the key as a Bearer token in the Authorization header.

For local use, run the package with npx and provide the key through AMT_API_KEY. The documented Claude command registers the local process over stdio. The same key format, beginning with amt_live_, is used for authentication in both deployment options.

Tools and capabilities

The core tool set includes:

• get_me, get_current_date, list_brands, and list_questions for workspace context and date-aware analysis.
• get_visibility, get_share_of_voice, and get_sentiment for brand measurements.
• list_competitors, list_answers, and list_sources for evidence, alternatives, citations, and domains.
• list_alerts and list_recommendations for findings and suggested next actions.

list_recommendations includes the pages and answers supporting each item. Check its displayable value: false means the available corpus is too small to rank recommendations. list_search_console is exposed only when the Google Search Console integration is enabled.

Limitations and notes

The AIMentionTracker MCP server does not provide write operations in version one. Adding prompts or triggering runs is outside the available tool set. Results also depend on the questions and answer volume configured for the workspace, so a disappointing metric may reflect insufficient or unsuitable question coverage.

AIMentionTracker should be used for answer-engine visibility analysis, not conventional rank tracking. Shares from workspaces with different numbers of tracked brands are not directly comparable, and sentiment counts should be interpreted with their classified denominator. The hosted service forwards the supplied key to the API for authentication, tenancy, entitlement, and rate limiting.