Skip to main content
mcp-geo product mark
Open source · MIT licensed

mcp-geo — the AI visibility MCP server.

Track ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews citations of your brand. MIT licensed. Self-host on your own Cloudflare account.

Worker / local tools
6 / 12
AI engines
7
Setup time
~5 min
License
MIT
mcp-geo AI visibility tracking overview: track AI citations, measure visibility, and monitor performance
mcp-geo connects AI citations, visibility measurement, and performance monitoring in one transparent workflow.

About

What is mcp-geo?

mcp-geo is an open-source Model Context Protocol (MCP) server for tracking how your brand is cited by AI search assistants. It runs prompts against ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews, then stores the raw citations and computed visibility metrics so you can query them from any MCP client. Think of it as the GEO (generative engine optimization) and AEO (answer engine optimization) equivalent of Google Search Console.

Run the npm package locally with SQLite for all twelve tools, including brand and prompt management, or self-host the six analysis tools on Cloudflare Workers + D1 for scheduled refreshes and remote access. You bring only the engine API keys you want to use. The software is free, but fresh scans may incur separate AI provider charges. Costs depend on the engines, prompts, model usage, and refresh frequency; see the cost breakdown.

Built for indie SaaS founders, solopreneurs, and small marketing teams who want to track AI visibility without paying for a bloated enterprise SEO suite. The full docs live in the README and the step-by-step SETUP.md on GitHub.

What it produces

Connect mcp-geo to Claude, ask it to run an AI visibility analysis on your brand, and within 90 seconds you get a strategist-quality memo grounded in per-engine data. The example below illustrates the report format.

Example report format. The brand shown is not presented as a customer or partner.

The full report continues below with content gaps, engine breakdown, and synthesis. Click the image above to view it all.

Features

What's inside

The same six analysis capabilities are available on both transports, but the tool names differ: Workers use the visibility.* namespace, while local stdio uses flat names. The local npm CLI also includes track_brand, list_brands, list_prompts, generate_prompts, update_brand, and set_prompts for twelve tools in total.

01
Worker
visibility.check
Local stdio
check_visibility

Latest AI visibility snapshot across all configured engines — per-engine scores, winning prompts, and losing prompts.

02
Worker
visibility.history
Local stdio
get_visibility_history

Time-series history of overall and per-engine visibility, bucketed daily or weekly.

03
Worker
visibility.compare
Local stdio
compare_competitors

Share-of-voice comparison against competitor domains, with prompts you win and prompts they win.

04
Worker
visibility.citations
Local stdio
get_citations

The actual citation events — prompt, engine, response excerpt, citation type, brand URL when present.

05
Worker
visibility.content_gaps
Local stdio
get_content_gaps

Prioritized content recommendations targeting your losing prompts.

06
Worker
visibility.refresh
Local stdio
refresh_brand

Manually trigger a fresh scan across every engine whose API key is set.

Engines

Supported engines

  • ChatGPT
  • Claude
  • Perplexity
  • Gemini
  • Google AI Overviews
  • Grok
  • Google AI Mode

Each engine is opt-in. You only configure the API keys you actually have — the worker simply skips engines whose credentials aren't set. The recommended starting pair is OpenAI + Anthropic; Gemini's API free tier has model-specific limits, and Google AI Overviews uses SerpApi search credits. Add engines deliberately once you want more coverage. Grok requires its own xAI key; Google AI Mode requires SerpApi and explicit SERPAPI_AI_MODE_ENABLED=true opt-in. Implementation details for every engine are in the README.

Costs and limits

Free MCP software. Variable external API costs.

Software

mcp-geo is free under the MIT license. Run the npm package locally with SQLite without paying DigestSEO for an MCP subscription.

Infrastructure

Self-hosting on Cloudflare Workers, D1 and Workers KV can fit within their limited free tiers. Limits apply to requests, compute, database reads/writes, KV operations and storage; free hosting is not unlimited.

AI providers

Fresh scans use your own API keys. Providers set their own prices, billing rules and free quotas. Charges can include input/output tokens, web searches or search credits; DigestSEO does not set those fees. Engines without keys are skipped.

For scale, a brand with 20 tracked prompts and two enabled engines attempts up to 40 provider requests per refresh. Weekly refreshes would attempt about 160–200 requests in a typical month, before setup-time prompt generation, retries or other provider calls. This is an illustrative request count, not a price quote. Token counts, search usage and billing depend on the provider and responses. A ChatGPT or Claude chat subscription does not include usage of these separate developer APIs.

Check current official pricing and limits before using any provider: OpenAI, Anthropic, Perplexity, Google Gemini, xAI, and SerpApi. For hosting, see Workers and D1 and Workers KV.

Setup

Run it locally with npm

The local CLI keeps its SQLite data on your machine and adds three brand-management tools. Set at least one supported engine API key; engines without a key are skipped.

claude mcp add --transport stdio digestseo -s user --env OPENAI_API_KEY=your_key_here -- npx -y @digestseo/mcp-geo

Or connect your Cloudflare Worker to Claude.ai

  1. In Claude.ai → Settings → Connectors → Add custom connector.
  2. Paste your worker URL after deploy (looks like https://YOUR-WORKER.YOUR-SUBDOMAIN.workers.dev/mcp).
  3. Complete the OAuth handshake. With the recommended CONNECT_SECRET set, enter that secret in the browser before a token is issued. Without it, anyone who knows your Worker URL can connect.
  4. Ask Claude: "Check the AI visibility for brand_id acme."

Or connect the Worker to Claude Code

claude mcp add --transport http digestseo https://YOUR-WORKER.YOUR-SUBDOMAIN.workers.dev/mcp

Claude Desktop, Cursor, and Codex CLI can use either the local npm package or connect to the Worker through mcp-remote — copy-pasteable configs for each client are in the README.

Deploy

5-minute self-host on Cloudflare

~/mcp-geo · deploy
# Full walkthrough with explanations: SETUP.md
git clone https://github.com/AKzar1el/mcp-geo.git
cd mcp-geo
npm install
npx wrangler login
cp wrangler.example.jsonc wrangler.jsonc

# Create KV + D1, paste the printed ids into wrangler.jsonc
npx wrangler kv namespace create OAUTH_KV
npx wrangler d1 create mcp-geo-db

# Required secret + at least one engine API key
npx wrangler secret put SEED_SECRET
npx wrangler secret put CONNECT_SECRET   # Recommended: gates who can connect
npx wrangler secret put OPENAI_API_KEY      # ChatGPT engine
npx wrangler secret put ANTHROPIC_API_KEY   # Claude engine + prompt generation

# Apply migrations and deploy
npx wrangler d1 migrations apply mcp-geo-db --remote
npx wrangler deploy

Done-for-you option

AI Visibility Audit — €99 one time

Prefer a client-ready baseline without running the stack yourself? The audit covers one brand/domain, up to 3 competitors, 20 buyer-intent prompts, and up to 5 supported AI surfaces where usable, with visibility/share-of-voice, citation evidence where returned, and a prioritized action memo.

Target delivery is within 2 business days after usable inputs. The open-source MCP remains free.

DigestSEO workspace

Start with a hosted workflow.

Create an account to try the DigestSEO workspace and keep your Site Audit projects in one place. For AI visibility tracking, self-host mcp-geo on Cloudflare.

Star the repo. Self-host in 5 minutes.

MIT licensed. 7 opt-in AI engines. Native MCP integration.