Daily Use Guide

A practical, copy-paste guide for auditing pages and domains every day with the Glippy MCP server, from Claude, ChatGPT, and Copilot, including how to run audits on a schedule.

Glippy scores how ready a site is for AI crawlers, LLM-powered search, and agents. The analysis runs locally on your machine; only the license check calls out to Glippy.

The 9 tools at a glance

You rarely name tools directly; you ask in plain language and the assistant picks the tool. This table is so you know what is possible.

Scoring in one line: each of the 16 categories scores 0-100, the overall is a weighted average, and grades map A+ (90+), A (80+), B (70+), C (60+), D (40+), F (under 40). See GEO Categories for the weights.

Set up your platform

You need a license key (GLMCP-XXXX-XXXX-XXXX) from glippy.dev/mcp, Node.js 18+, and optionally Chrome for the anti-bot fallback. For full setup per environment, see Getting Started and Integrations. Highlights per platform:

Claude (recommended, native)

Add Glippy to Claude Code (.mcp.json) or Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "glippy-geo": {
      "command": "npx",
      "args": ["-y", "glippy-mcp"],
      "env": { "GLIPPY_LICENSE_KEY": "GLMCP-XXXX-XXXX-XXXX" }
    }
  }
}

Restart the client, then confirm with /mcp in Claude Code, or the tools icon in Claude Desktop.

GitHub Copilot (VS Code agent mode)

Copilot's agent mode speaks MCP over stdio, so the same local server works. Create .vscode/mcp.json in your workspace:

{
  "servers": {
    "glippy-geo": {
      "command": "npx",
      "args": ["-y", "glippy-mcp"],
      "env": { "GLIPPY_LICENSE_KEY": "GLMCP-XXXX-XXXX-XXXX" }
    }
  }
}

Open Copilot Chat, switch it to Agent mode, and Glippy's tools appear in the tools list.

ChatGPT (workaround today, native app coming)

ChatGPT does not run local stdio MCP servers, so it cannot call Glippy out of the box. Three honest options:

• Wait for the official Glippy ChatGPT app (in development; sign up at glippy.dev).
• Run an MCP-to-OpenAI bridge locally so your own OpenAI-based agent can call the tools:

  npx mcp-openai-bridge --server "npx -y glippy-mcp"

• Paste and interpret: run the audit in Claude or Copilot, export a report, then paste it into ChatGPT and ask it to prioritise or draft the fixes. ChatGPT is excellent at the "what do I do about this" step even when it cannot run the crawl.

Daily workflows

These are natural-language prompts. They work the same in Claude and in Copilot agent mode. Swap in your domain.

The 30-second morning check

Give me a quick GEO summary of glippy.dev. Use render_mode auto.

Returns overall score, grade, top 3 strengths, and top 3 issues. Perfect for a daily glance.

Full single-domain audit

Run a full GEO analysis on glippy.dev, crawl up to 5 pages, render_mode auto.
Then list the High-priority issues first.

Use this when you actually plan to fix things. max_pages is 1-10; more pages is slower but catches template-wide problems.

Competitor benchmark

Compare GEO readiness of glippy.dev, competitor1.com, and competitor2.com.
Rank them and show the category comparison table.

Great as a weekly ritual; tells you exactly which categories you are losing on.

Section or whole-site sweep

Analyse the pages in https://glippy.dev/sitemap.xml, cap it at 100 URLs,
rate_limit 2. Show me the 5 weakest pages and the most common issues.

Best for finding template-level problems that repeat across many pages.

Specific page batch

Audit these pages: https://glippy.dev/pricing, https://glippy.dev/docs,
https://glippy.dev/blog/geo-guide. Which one is weakest and why?

Make a shareable report

Generate an HTML GEO report for glippy.dev and save it as glippy-report.html.

Formats: html (styled standalone page), markdown (recommendations only), markdown_full (every check). For multiple domains or a sitemap, ask for a "bulk report".

AI crawler access spot-check

Which AI crawlers does glippy.dev block in robots.txt?

Checks GPTBot, ClaudeBot, Google-Extended, PerplexityBot, CCBot, Bytespider, and more, plus any Content-Signal directive.

Time-saving tips

• Big sites returning empty or zero scores are almost always bot-blocked, not broken. Sites behind Cloudflare, Akamai, or DataDome refuse plain fetches. Add render_mode auto (static first, Chrome fallback only when blocked) or render_mode chrome (always Chrome). Check the renderMode field in the result to see which path ran.
• Analyse once, export many. Ask for output_format json on the analysis, then hand that JSON to export_report or export_bulk_report. Results are also cached in memory for 5 minutes, so a follow-up export usually will not re-crawl.
• Be polite on sites you do not own. Default rate_limit is 5 req/sec per domain. Drop to 0.5-1 for third-party sites; raise to 10-50 only on infrastructure you control.
• Read grades with the weights in mind. Structured Data and Machine Readability (both 1.5x) and Citability (1.3x) move the overall score most. Performance & Crawlability (0.3x) and Agent Interactivity (0.2x) are deliberately low priority, so do not over-invest there.
• llms.txt is not a real ranking factor today. No major model or crawler consumes it yet. A missing llms.txt is not a problem worth prioritising; treat it as nice to have.
• Partial hreflang is usually fine. On a site that is only partly translated, incomplete hreflang coverage is expected and not automatically a fix-it-now issue.
• Emerging agent surfaces are bonus-scored. MCP server cards, A2A cards, NLWeb, and schemamap earn credit when present but are not penalised when absent, so a low Agent Interactivity score is not an emergency.
• Auditing your own site? Make sure your WAF or robots.txt is not blocking the GlippyBot user-agent. See the GlippyBot allow-list guide.

Schedule recurring audits

Because Glippy is a local stdio program, only tools that run on your machine or in CI can drive it on a timer. Here is the honest layout.

Option 1: Claude Code routines (simplest native path)

Inside Claude Code, ask it to schedule a recurring agent that runs your audit prompt:

Every weekday at 08:00, run a full GEO analysis on glippy.dev with render_mode auto,
export an HTML report to ./reports/, and summarise any new High-priority issues.

Claude Code registers a cron routine that fires on schedule, calls the Glippy MCP tools, and leaves the report in ./reports/.

Option 2: OS scheduler + Claude Code headless (most portable)

Claude Code can run non-interactively with -p (print mode), which makes it scriptable. On Windows (PowerShell), register a daily task:

$action  = New-ScheduledTaskAction -Execute "claude" `
  -Argument '-p "Run get_geo_summary on glippy.dev with render_mode auto and append the score to reports\daily-scores.md"'
$trigger = New-ScheduledTaskTrigger -Daily -At 8:00am
Register-ScheduledTask -TaskName "Glippy daily audit" -Action $action -Trigger $trigger

The equivalent on macOS or Linux is a cron line calling the same claude -p "..." command. Make sure the GLIPPY_LICENSE_KEY env var and your .mcp.json are available to the scheduled shell.

Option 3: GitHub Actions (scheduled, in CI)

Good for auditing a site on a schedule and committing the report or failing the build if the score drops:

name: Daily GEO audit
on:
  schedule:
    - cron: "0 8 * * *"   # 08:00 UTC daily
  workflow_dispatch:
jobs:
  audit:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with: { node-version: 20 }
      - name: Run Glippy audit
        env:
          GLIPPY_LICENSE_KEY: ${{ secrets.GLIPPY_LICENSE_KEY }}
        run: |
          # drive the MCP via your preferred runner (Claude Code headless,
          # the mcp-openai bridge, or a small Node script)
          npx -y glippy-mcp --help || true

Use render_mode static in CI unless you also install Chrome on the runner.

Option 4: ChatGPT Tasks and Copilot Studio (the interpret layer)

• ChatGPT Tasks can run a prompt on a schedule, but a Task cannot reach your local Glippy server. Use it for the human-facing half: have an OS job or GitHub Action write the report somewhere, then let a scheduled ChatGPT Task read and interpret it and message you the priorities.
• Copilot Studio / Power Automate can trigger agents on a schedule, but only against remote MCP servers, so it cannot drive Glippy until a hosted endpoint exists. GitHub Actions (Option 3) is the Microsoft-friendly stand-in.

Prompt library

Paste and go. Replace DOMAIN, RIVAL1, and the URLs with your own.

Quick check:    Give me a quick GEO summary of DOMAIN with render_mode auto.
Full audit:     Full GEO analysis of DOMAIN, 5 pages, render_mode auto, High-priority issues first.
Competitors:    Compare GEO readiness of DOMAIN, RIVAL1, RIVAL2. Rank and show the category table.
Whole site:     Analyse https://DOMAIN/sitemap.xml, max 100 URLs, rate_limit 2. Show weakest pages and common issues.
Page batch:     Audit URL1, URL2, URL3. Which is weakest and why?
Crawler access: Which AI crawlers does DOMAIN block in robots.txt?
Share it:       Generate an HTML GEO report for DOMAIN and save it.
Fix plan:       From that audit, write a prioritised fix list a developer can action this sprint.

Next Steps