MCP Tools Reference

Complete reference for all SeeLLM MCP tools available to AI assistants.

Contract

SeeLLM's MCP server exposes seven read-only tools for the authenticated organization. Machines can use the versioned MCP tools manifest, which includes the JSON Schema inputs, output shape, authentication requirement, errors, and a link back to this reference.

AI-related requests are HTTP requests classified into assistant, training-crawler, coding-agent, or AI-referral categories. They are not unique people or sessions.

Every successful call returns this MCP result shape. text is Markdown, with the headings and tables noted for each tool below.

{
  "content": [{ "type": "text", "text": "# Result heading\n\n..." }]
}

All tools require Authorization: Bearer sk_live_…; the MCP endpoint accepts active SeeLLM API keys, not Firebase JWTs. They return data only for the authenticated organization. Missing, invalid, expired, or non-API-key credentials return HTTP 401 Unauthorized. MCP rejects input that does not match the documented schema before the handler runs; data-retrieval failures are surfaced as MCP tool errors.

Available Tools

get_traffic_overview

Gets observed HTTP request totals, the human/AI-related split, the four AI-related request categories, and the top AI-related request sources.

Output: # Traffic Overview followed by a table containing Total Visits, Human Visits, AI-related Requests, Assistant Requests, Training Crawler Requests, Coding Agent Requests, AI Referral Requests, Other Automated, and AI-related Request Share. The top-level counts reconcile: total equals human plus AI-related plus other automated. If data exists, it also includes an ## AI-related Request Sources table (up to 10 sources). These request counts do not establish citations or visits by unique people.

Example prompt: “Show me my traffic overview for the last 7 days.”

get_ai_platform_breakdown

Gets the observed assistant-only request mix by platform. It excludes model-training crawlers, ordinary search crawlers, and unknown automation. This is narrower than all AI-related requests and does not prove citation.

Output: # AI Assistant Platform Breakdown and a Platform/AI Assistant Requests/Share/Pages Accessed/Unique Visitors table. Pages Accessed counts distinct non-empty requested paths. Unique Visitors is a compatibility label for distinct available privacy-safe visitor hashes, not verified unique people or sessions; — means no such identifier was available for the selected data. With no rows, the Markdown includes a no-data notice.

Example prompt: “Which AI platforms are visiting my site most?”

get_top_pages

Gets meaningful content pages ranked by canonical human and AI requests for one domain. It excludes monitoring, metadata, static assets, and malicious probe paths.

Output: # Top Pages and a ranked Path/Requests table over canonical human and AI requests, or a no-data notice.

Example prompt: “What are the most requested monitored pages on example.com?”

get_content_valuation

Estimates the monetary value of content accessed by AI platforms for one domain.

Output: # Content Valuation with the detected vertical, base CPM, total estimated value, and total crawls. It includes ## By Platform and ## By Content Type tables, or a no-data notice when both are empty.

Example prompt: “What is the estimated value of AI-accessed content on example.com?”

get_alerts

Gets recent AI citation alerts for the authenticated organization.

Output: # AI Alerts followed by one Markdown bullet per alert with severity, type, platform, optional domain, status, and timestamp. It returns a no-alerts notice when empty.

Example prompt: “Show me my open AI citation alerts.”

whatif_block_platform

Estimates the referral and value impact of blocking an AI platform for one domain.

Output: # What-If: Block … and a table of current crawls, referrals, projected referral loss, loss percent, and revenue impact, followed by a recommendation. With no platform data, it returns the corresponding safety message.

Example prompt: “What would happen if I blocked chatgpt from example.com?”

get_weekly_trend

Gets a weekly request trend with human, AI-related, and other-automated totals plus optional per-platform detail.

Output: # Weekly Traffic Trend and a Week/Total/Human/AI-related Requests/Other Automated/AI-related Request Share table. Each weekly total reconciles across the three traffic classes. When platform data exists, it adds ## AI-related Requests by Platform for up to eight named platforms across assistant, training-crawler, coding-agent, and AI-referral requests. AI-related requests with unknown sources are omitted from that detail table.

Example prompt: “Show the weekly AI traffic trend for example.com over 90 days.”