SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI developer platformscitation-intelligence

citation-intelligence

Check which URLs Perplexity, Claude, ChatGPT, Gemini, Bing and Google AI Overviews cite for any query. Self-hosted.

Open website

About this tool

Purpose

Citation Intelligence is an MCP server that helps agents and developers understand which URLs will be cited by AI search engines for specific queries. It covers Perplexity, Google AI Overviews, ChatGPT, Claude, Gemini, and Bing. The goal is to let citation data flow directly into your workflow, rather than into some SaaS login screen.

Main capabilities

Tools are grouped into seven namespaces:

  • citations_*: query level—who cited what, with what evidence.
  • domain_*: domain level—am I cited, and for what.
  • signals_*: external signals—AI Overview, Wikipedia, GSC, answer box position.
  • panel_*: saved query panels (edit watchlists).
  • report_*: out-of-the-box report artifacts.
  • competitors_*: competitive landscape by query.
  • audit_*: fixable on-page/on-site checks.

Server-side prompt templates include: audit_citation_readiness(url), audit_competitor_snapshot(query, your_url?), audit_crawler_checkup(url), audit_gap_analysis(domain, days?), audit_sitemap_coverage(sitemap_url).

Cached views that clients can read or subscribe to (no tool call required): citation://cache/summary, citation://panels, citation://docs/llms-txt, citation://docs/ai-crawlers, citation://domain/{domain}/cited-for.

Measurement method and surface fidelity

Every response includes a surface field explaining how the data was collected:

  • perplexity (consumer_scrape): via the Perplexity API using Sonar Pro and a near-consumer-grade system prompt; results are relatively close to Perplexity.ai.- claude (api_proxy): via the Anthropic Messages API using Claude Sonnet with the web_search tool enabled; consumer Claude.ai uses different routing and ranking logic, so citation behavior may differ.
  • openai (api_proxy): via the OpenAI Responses API using gpt-4o + the web_search_preview tool.
  • gemini (api_proxy): via the Generative Language API using Gemini 2.5 Pro with google_search grounding enabled; consumer Gemini uses the same grounding index but reranks differently, so results are only directional.
  • google_ai_mode (consumer_scrape): retrieves Google AI Mode results via SerpAPI, closest to what users see in Google Search, requires SERPAPI_KEY.
  • bing_serp / brave_serp (web_rank): traditional SERP rankings, does not measure LLM citations; domain_am_i_cited rejects these engines and measures only LLM behavior.

Every tool response includes an interpretation_note field that summarizes fidelity in one sentence.

Usage

Requires Node 20 or later. Add the server to your MCP client configuration (on Windows, %APPDATA%\Claude\claude_desktop_config.json; on macOS, ~/Library/Application Support/Claude/claude_desktop_config.json). Set only the keys you own. Any MCP client that supports stdio transport can use the same command/argument pattern.

Environment variables are used to provide API keys; keys are read from the MCP process environment, never logged, and never persisted. The cache file is located at ~/.config/citation-intelligence/cache.json and can bedeleted, default TTL is 7 days.

Use cases

  • Build research agents that cite sources the LLM already trusts.
  • Independent developers or indie hackers check whether their own SaaS appears in AI search.
  • Content creators confirm whether articles are cited by ChatGPT, Claude, or Perplexity.
  • SEO or GEO practitioners want to obtain citation data programmatically without relying on a $295-499 per month dashboard.
  • Editorial workflows want to choose topics based on citation gaps.
  • Compare competitors' visibility across AI engines in any niche.

Limitations and non-use cases

  • Not suitable for users who need polished marketing dashboards, charts, and team seats.
  • Not a hosted service with an SLA; self-hosted by design.
  • Not for academic paper citation tracking.
  • Does not include 350 million+ pre-modeled prompts.

How it stays free: no central backend, the server runs on your machine, nothing is uploaded; prioritize free quotas (SerpAPI offers 100 free Google AI Overview queries per month, Bing Web Search has a free tier, Perplexity offers free Sonar access after signup); bring your own paid keys to enable premium engines; local caching makes repeated queries hit the cache instead of the API; citations_predict runs without any keys; all API calls go directly from your machine to the vendors; no proxy, no analytics, no telemetry by default.

Workflow examples

  • Weekly citation tracking: select 20-30 queries from the editorial backlog, snapshot weekly, and observe ratio trends; citations_trend returns the delta for each query.
  • Pre-publication gate: check who owns the citation slot and whether that slot is worth competing for.
  • Batch site audit: check site-level structural issues across all pages at once, with zero API overhead.
  • Competitor signal gaps: view diverging_signals to find where you are falling behind.
  • Google rankings vs. AI citation gaps: requires a GCP service account with the webmasters.readonly scope; closest_wins returns queries with position <= 10 and ai_cited: false.
  • Wikipedia mention monitoring: rerun quarterly, and the diff is the "wegot a Wikipedia citation" alert.

The license is MIT, built by automatelab.tech.

Related tools

Browse category