@fullmention/mcp-server

The FullMention Model Context Protocol (MCP) Server is a secure, high-performance gateway that allows Large Language Models (LLMs) and AI agents (such as Claude Desktop, Cursor, or custom agent networks) to interact directly with the FullMention Public API.

By exposing FullMention’s stateless recommendation snapshots as native MCP tools and structured resources, your AI models can seamlessly monitor AI visibility, manage keywords, estimate credit costs, and trigger batch analysis runs.

🚀 Quickstart: Connect your AI Client (Local Desktop)

Since the server is published as a zero-install NPM package, you can connect it instantly using npx.

1. Claude Desktop Setup

Add the following configuration to your claude_desktop_config.json (typically located at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "fullmention": {
      "command": "npx",
      "args": ["-y", "@fullmention/mcp-server"],
      "env": {
        "FULLMENTION_API_KEY": "your_fm_live_or_test_key_here"
      }
    }
  }
}

2. Cursor Setup

1. Open Cursor Settings and navigate to Features -> MCP.

2. Click + Add New MCP Server.

3. Fill in the server parameters:

   • Name: FullMention

   • Type: stdio

   • Command: npx -y @fullmention/mcp-server

4. Add the environment variable FULLMENTION_API_KEY and paste your key.

5. Click Save.

🛡️ Core Architecture & Smart Shielding

Designed from the ground up to be AI-safe and highly resource-efficient, automatically shielding your API quotas and credit budgets:

• 🛡️ In-Memory Query Shielding: Caches query responses (60-second TTL) to protect your external API rate limits from redundant agent loops.

• 🚦 Status Polling Guardrails: Detects rapid-succession polling of active batch-run statuses. If an AI agent polls the server within 8 seconds, the server intercepts the network call and returns cached state with a [POLLED_TOO_FAST_CACHED_WARNING], saving network overhead and API stress.

• ⚡ Conflict Safety: Gracefully handles 409 Conflict scenarios if a run is already active under specific tag filters, automatically extracting the active runId and returning [RUN_ALREADY_PROCESSING_CONFLICT] to guide the AI model without crashing.

• 📊 Server-Side Share of Voice: Automates brand-visibility calculation server-side with get_share_of_voice to completely prevent mathematical hallucinations in LLMs.

• 🚀 Server-Side Token Optimization: Supports a format: "markdown" parameter. Instead of sending raw verbose JSON to the LLM, the MCP server automatically compiles your data into highly compressed, dense Markdown tables, saving massive token overhead and drastically reducing your LLM input cost.

🌐 Exponentiate with Cloud Deployment (SSE Mode)

For SaaS and cloud architectures, the FullMention MCP Server supports SSE (Server-Sent Events) HTTP Transport dynamically out of the box.

Launching in SSE Mode:

To spin up the HTTP gateway on port 3000 (or configured PORT / FULLMENTION_MCP_PORT):

npx -y @fullmention/mcp-server --sse

Read-Only Security Mode

For public gateways or shared team environments, you can configure the MCP server in a strictly read-only mode by passing FULLMENTION_MCP_READONLY=true. This blocks all mutating actions (such as triggering runs, creating or deleting keywords):

PORT=3000 FULLMENTION_MCP_READONLY=true npx -y @fullmention/mcp-server --sse

100% Secure Metrics Endpoint

When hosted in SSE mode, you can inspect diagnostics and live caching performance metrics via:

curl http://localhost:3000/diagnostics

📋 Detailed Tool & Parameter Reference

FullMention MCP server exposes 17 dedicated tools. Below is the complete parameter schema and response shape for each tool:

📊 System & Quota Tools

1. get_status

Retrieve dynamic system and search engine status.

• Parameters: None.

• Return (JSON):

  {
    "status": "operational",
    "engines": {
      "openai": "operational",
      "gemini": "operational"
    }
  }

2. get_quota

Retrieve account quota limits and current usage.

• Parameters:

  • bypassCache (boolean, optional): If true, bypasses in-memory caching.

• Return (JSON):

  {
    "monthlyLimit": 10000,
    "remaining": 7450,
    "reserved": 200,
    "used": 2350
  }

3. get_usage_stats

Retrieve detailed credit usage stats and pricing matrix.

• Parameters:

  • bypassCache (boolean, optional): If true, bypasses in-memory caching.

• Return: Detailed Markdown summary including remaining credits, monthly limit, and the active cost matrix.

🗃️ Keyword Management

4. list_keywords

List all monitored keywords in FullMention with pagination.

• Parameters:

  • tags (string, optional): Comma-separated tags to filter by (e.g., "client:acme,market:dk").

  • tagMode ("and" | "or", optional, default: "or"): Match mode for tag filter.

  • country (string, optional): Country name filter (e.g., "Denmark").

  • language (string, optional): Language name filter (e.g., "Danish").

  • location (string, optional): City level location (e.g., "Copenhagen").

  • engine ("openai" | "openai-mini" | "gemini", optional): AI search engine.

  • limit (number, optional, default: 100): Limit keywords per page.

  • cursor (string, optional): Cursor for pagination.

  • format ("compact" | "raw" | "markdown", optional, default: "compact"): Return payload format.

  • bypassCache (boolean, optional): If true, pulls live data.

5. create_keywords

Create one or more keywords idempotent.

• Parameters:

  • keywords (array of objects, required):

    • keyword (string, required): Search query intent (e.g., "best running shoes").

    • country (string, required): Country name (e.g., "Denmark").

    • countryCode (string, required): ISO 2-letter code (e.g., "DK").

    • language (string, required): Language name (e.g., "Danish").

    • languageCode (string, required): ISO 2-letter code (e.g., "da").

    • location (string, optional, nullable): City/local level (e.g., "Copenhagen").

    • engines (array of strings, optional, default: ["openai-mini"]).

    • tags (array of strings, optional): Prefixed tags.

  • idempotencyKey (string, optional): Unique UUID to prevent double-creation.

6. bulk_create_keywords

Bulk-create keywords using a simple string array with shared geographical parameters.

• Parameters:

  • keywords (array of strings, required): Intents to create.

  • country / countryCode / language / languageCode / location (optional): Shared config.

  • engines (optional, default: ["openai-mini"]): Target AI engines.

  • tags (optional): Shared tags.

  • idempotencyKey (string, optional).

7. update_keyword

Update settings or tags for an existing keyword.

• Parameters:

  • keywordId (string, required): The target keyword ID.

  • keyword / country / countryCode / language / languageCode / location / engines / tags (optional).

  • idempotencyKey (string, optional).

8. delete_keyword

Deactivate or soft-delete a keyword.

• Parameters:

  • keywordId (string, required): Keyword ID.

  • idempotencyKey (string, optional).

🚦 Tags & Runs (Batch Analysis)

9. list_tags

List active tags currently utilized across your keywords.

• Parameters:

  • prefix (string, optional): Prefix filter (e.g., "client:").

  • limit / cursor / bypassCache (optional).

10. estimate_run

Pre-estimate credit consumption prior to starting a batch analysis.

• Parameters:

  • tags (array of strings, required): Tag filters determining keywords.

  • tagMode ("and" | "or", optional): Match mode.

  • fanout (boolean, optional): Enable web search search.

• Return (JSON):

  {
    "estimatedCredits": 25,
    "keywordCount": 25,
    "insufficientCredits": false
  }

11. trigger_run

Trigger an asynchronous batch analysis (run) on all matching keywords.

• Parameters:

  • tags (array of strings, required): Target tags.

  • tagMode ("and" | "or", optional).

  • fanout (boolean, optional): Enabling search query fanout costs extra credits.

  • idempotencyKey (string, optional).

12. get_run_status

Fetch progress and active status of a batch run.

• Parameters:

  • runId (string, required): Run ID.

13. cancel_run

Cancel an active batch analysis run and release remaining reserved quota.

• Parameters:

  • runId (string, required): Run ID.

14. list_runs

List recent batch runs with timestamps and status logs.

• Parameters:

  • limit / cursor / bypassCache (optional).

📈 Results & Analytics

15. get_latest_results

Retrieve the latest AI search visibility recommendations.

• Parameters:

  • tags (string, optional): Comma-separated tag filter.

  • tagMode / country / language / location / engine / keywordId (optional).

  • limit / cursor / bypassCache (optional).

  • format ("compact" | "raw" | "markdown", optional, default: "compact").

16. get_share_of_voice

Calculate brand Share of Voice % and average placements server-side.

• Parameters:

  • tags (string, required): Target tags for calculation.

  • tagMode ("and" | "or", optional).

  • engine (optional): Specific search engine filter.

  • brands (array of strings, optional): Brands to isolate.

  • format ("json" | "markdown", optional, default: "markdown").

  • bypassCache (boolean, optional).

• Return: Compiles a dense Markdown table reporting visibility share percentages and rank metrics.

17. get_fanout_sources

Retrieve the underlying web query search sources that influenced a recommendation.

• Parameters:

  • resultId (string, required).

  • limit / cursor (optional).

🧩 Exposed Native Resources

Your AI models can consume structured datasets directly using these resource URIs:

• fullmention://keywords/active — Real-time markdown table containing all active keywords.

• fullmention://schema/openapi — The complete machine-readable OpenAPI spec.

• fullmention://results/latest/tag/{tag} — Live markdown recommendation summaries for a given tag.

⚠️ Known Issues

• Engine Filter 500 Bug: Filtering results or list endpoints strictly by engine (specifically gemini since Gemini does not support web search fanout) can occasionally return a 500 Internal Server Error from the backend API.

  • Workaround: Omit the engine parameter to fallback safely to the default openai-mini pipeline.